You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
minio/cmd/admin-heal-ops.go

880 lines
24 KiB

Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
/*
* MinIO Cloud Storage, (C) 2017 MinIO, Inc.
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package cmd
import (
"context"
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
"encoding/json"
"fmt"
"net/http"
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
"strings"
"sync"
"time"
"github.com/minio/minio/cmd/logger"
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
"github.com/minio/minio/pkg/madmin"
)
// healStatusSummary - overall short summary of a healing sequence
type healStatusSummary string
// healStatusSummary constants
const (
healNotStartedStatus healStatusSummary = "not started"
healRunningStatus = "running"
healStoppedStatus = "stopped"
healFinishedStatus = "finished"
)
const (
// a heal sequence with this many un-consumed heal result
// items blocks until heal-status consumption resumes or is
// aborted due to timeout.
maxUnconsumedHealResultItems = 1000
// if no heal-results are consumed (via the heal-status API)
// for this timeout duration, the heal sequence is aborted.
healUnconsumedTimeout = 24 * time.Hour
// time-duration to keep heal sequence state after it
// completes.
keepHealSeqStateDuration = time.Minute * 10
// nopHeal is a no operating healing action to
// wait for the current healing operation to finish
nopHeal = ""
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
)
var (
errHealIdleTimeout = fmt.Errorf("healing results were not consumed for too long")
errHealStopSignalled = fmt.Errorf("heal stop signaled")
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
errFnHealFromAPIErr = func(ctx context.Context, err error) error {
apiErr := toAPIError(ctx, err)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
return fmt.Errorf("Heal internal error: %s: %s",
apiErr.Code, apiErr.Description)
}
)
// healSequenceStatus - accumulated status of the heal sequence
type healSequenceStatus struct {
// summary and detail for failures
Summary healStatusSummary `json:"Summary"`
FailureDetail string `json:"Detail,omitempty"`
StartTime time.Time `json:"StartTime"`
// disk information
NumDisks int `json:"NumDisks"`
// settings for the heal sequence
HealSettings madmin.HealOpts `json:"Settings"`
// slice of available heal result records
Items []madmin.HealResultItem `json:"Items"`
}
// structure to hold state of all heal sequences in server memory
type allHealState struct {
sync.Mutex
// map of heal path to heal sequence
healSeqMap map[string]*healSequence
}
// initHealState - initialize healing apparatus
func initHealState() *allHealState {
healState := &allHealState{
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
healSeqMap: make(map[string]*healSequence),
}
go healState.periodicHealSeqsClean(GlobalContext)
return healState
}
func (ahs *allHealState) periodicHealSeqsClean(ctx context.Context) {
// Launch clean-up routine to remove this heal sequence (after
// it ends) from the global state after timeout has elapsed.
for {
select {
case <-time.After(time.Minute * 5):
now := UTCNow()
ahs.Lock()
for path, h := range ahs.healSeqMap {
if h.hasEnded() && h.endTime.Add(keepHealSeqStateDuration).Before(now) {
delete(ahs.healSeqMap, path)
}
}
ahs.Unlock()
case <-ctx.Done():
// server could be restarting - need
// to exit immediately
return
}
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
// getHealSequenceByToken - Retrieve a heal sequence by token. The second
// argument returns if a heal sequence actually exists.
func (ahs *allHealState) getHealSequenceByToken(token string) (h *healSequence, exists bool) {
ahs.Lock()
defer ahs.Unlock()
for _, healSeq := range ahs.healSeqMap {
if healSeq.clientToken == token {
return healSeq, true
}
}
return nil, false
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// getHealSequence - Retrieve a heal sequence by path. The second
// argument returns if a heal sequence actually exists.
func (ahs *allHealState) getHealSequence(path string) (h *healSequence, exists bool) {
ahs.Lock()
defer ahs.Unlock()
h, exists = ahs.healSeqMap[path]
return h, exists
}
func (ahs *allHealState) stopHealSequence(path string) ([]byte, APIError) {
var hsp madmin.HealStopSuccess
he, exists := ahs.getHealSequence(path)
if !exists {
hsp = madmin.HealStopSuccess{
ClientToken: "invalid",
StartTime: UTCNow(),
}
} else {
hsp = madmin.HealStopSuccess{
ClientToken: he.clientToken,
ClientAddress: he.clientAddress,
StartTime: he.startTime,
}
he.stop()
for !he.hasEnded() {
time.Sleep(1 * time.Second)
}
ahs.Lock()
defer ahs.Unlock()
// Heal sequence explicitly stopped, remove it.
delete(ahs.healSeqMap, path)
}
b, err := json.Marshal(&hsp)
return b, toAdminAPIErr(GlobalContext, err)
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// LaunchNewHealSequence - launches a background routine that performs
// healing according to the healSequence argument. For each heal
// sequence, state is stored in the `globalAllHealState`, which is a
// map of the heal path to `healSequence` which holds state about the
// heal sequence.
//
// Heal results are persisted in server memory for
// `keepHealSeqStateDuration`. This function also launches a
// background routine to clean up heal results after the
// aforementioned duration.
func (ahs *allHealState) LaunchNewHealSequence(h *healSequence) (
respBytes []byte, apiErr APIError, errMsg string) {
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
existsAndLive := false
he, exists := ahs.getHealSequence(h.path)
if exists {
existsAndLive = !he.hasEnded()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
if existsAndLive {
// A heal sequence exists on the given path.
if h.forceStarted {
// stop the running heal sequence - wait for it to finish.
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
he.stop()
for !he.hasEnded() {
time.Sleep(1 * time.Second)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
} else {
errMsg = "Heal is already running on the given path " +
"(use force-start option to stop and start afresh). " +
fmt.Sprintf("The heal was started by IP %s at %s, token is %s",
h.clientAddress, h.startTime.Format(http.TimeFormat), h.clientToken)
return nil, errorCodes.ToAPIErr(ErrHealAlreadyRunning), errMsg
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
}
ahs.Lock()
defer ahs.Unlock()
// Check if new heal sequence to be started overlaps with any
// existing, running sequence
for k, hSeq := range ahs.healSeqMap {
if !hSeq.hasEnded() && (strings.HasPrefix(k, h.path) ||
strings.HasPrefix(h.path, k)) {
errMsg = "The provided heal sequence path overlaps with an existing " +
fmt.Sprintf("heal path: %s", k)
return nil, errorCodes.ToAPIErr(ErrHealOverlappingPaths), errMsg
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
}
// Add heal state and start sequence
ahs.healSeqMap[h.path] = h
// Launch top-level background heal go-routine
go h.healSequenceStart()
b, err := json.Marshal(madmin.HealStartSuccess{
ClientToken: h.clientToken,
ClientAddress: h.clientAddress,
StartTime: h.startTime,
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
})
if err != nil {
logger.LogIf(h.ctx, err)
return nil, toAPIError(h.ctx, err), ""
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
return b, noError, ""
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
// PopHealStatusJSON - Called by heal-status API. It fetches the heal
// status results from global state and returns its JSON
// representation. The clientToken helps ensure there aren't
// conflicting clients fetching status.
func (ahs *allHealState) PopHealStatusJSON(path string,
clientToken string) ([]byte, APIErrorCode) {
// fetch heal state for given path
h, exists := ahs.getHealSequence(path)
if !exists {
// If there is no such heal sequence, return error.
return nil, ErrHealNoSuchProcess
}
// Check if client-token is valid
if clientToken != h.clientToken {
return nil, ErrHealInvalidClientToken
}
// Take lock to access and update the heal-sequence
h.mutex.Lock()
defer h.mutex.Unlock()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
numItems := len(h.currentStatus.Items)
// calculate index of most recently available heal result
// record.
lastResultIndex := h.lastSentResultIndex
if numItems > 0 {
lastResultIndex = h.currentStatus.Items[numItems-1].ResultIndex
}
h.lastSentResultIndex = lastResultIndex
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
jbytes, err := json.Marshal(h.currentStatus)
if err != nil {
h.currentStatus.Items = nil
logger.LogIf(h.ctx, err)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
return nil, ErrInternalError
}
h.currentStatus.Items = nil
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
return jbytes, ErrNone
}
// healSource denotes single entity and heal option.
type healSource struct {
path string // entity path (format, buckets, objects) to heal
opts madmin.HealOpts // optional heal option overrides default setting
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// healSequence - state for each heal sequence initiated on the
// server.
type healSequence struct {
// bucket, and prefix on which heal seq. was initiated
bucket, objPrefix string
// path is just pathJoin(bucket, objPrefix)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
path string
// A channel of entities (format, buckets, objects) to heal
sourceCh chan healSource
// A channel of entities with heal result
respCh chan healResult
// Report healing progress
reportProgress bool
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// time at which heal sequence was started
startTime time.Time
// time at which heal sequence has ended
endTime time.Time
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// Heal client info
clientToken, clientAddress string
// was this heal sequence force started?
forceStarted bool
// heal settings applied to this heal sequence
settings madmin.HealOpts
// current accumulated status of the heal sequence
currentStatus healSequenceStatus
// channel signaled by background routine when traversal has
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// completed
traverseAndHealDoneCh chan error
// canceler to cancel heal sequence.
cancelCtx context.CancelFunc
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// the last result index sent to client
lastSentResultIndex int64
// Number of total items scanned against item type
scannedItemsMap map[madmin.HealItemType]int64
// Number of total items healed against item type
healedItemsMap map[madmin.HealItemType]int64
// Number of total items where healing failed against endpoint and drive state
healFailedItemsMap map[string]int64
// The time of the last scan/heal activity
lastHealActivity time.Time
// Holds the request-info for logging
ctx context.Context
// used to lock this structure as it is concurrently accessed
mutex sync.RWMutex
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
// NewHealSequence - creates healSettings, assumes bucket and
// objPrefix are already validated.
func newHealSequence(ctx context.Context, bucket, objPrefix, clientAddr string,
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
numDisks int, hs madmin.HealOpts, forceStart bool) *healSequence {
reqInfo := &logger.ReqInfo{RemoteHost: clientAddr, API: "Heal", BucketName: bucket}
reqInfo.AppendTags("prefix", objPrefix)
ctx, cancel := context.WithCancel(logger.SetReqInfo(ctx, reqInfo))
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
return &healSequence{
respCh: make(chan healResult),
bucket: bucket,
objPrefix: objPrefix,
path: pathJoin(bucket, objPrefix),
reportProgress: true,
startTime: UTCNow(),
clientToken: mustGetUUID(),
clientAddress: clientAddr,
forceStarted: forceStart,
settings: hs,
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
currentStatus: healSequenceStatus{
Summary: healNotStartedStatus,
HealSettings: hs,
NumDisks: numDisks,
},
traverseAndHealDoneCh: make(chan error),
cancelCtx: cancel,
ctx: ctx,
scannedItemsMap: make(map[madmin.HealItemType]int64),
healedItemsMap: make(map[madmin.HealItemType]int64),
healFailedItemsMap: make(map[string]int64),
}
}
// resetHealStatusCounters - reset the healSequence status counters between
// each monthly background heal scanning activity.
// This is used only in case of Background healing scenario, where
// we use a single long running healSequence which reactively heals
// objects passed to the SourceCh.
func (h *healSequence) resetHealStatusCounters() {
h.mutex.Lock()
defer h.mutex.Unlock()
h.currentStatus.Items = []madmin.HealResultItem{}
h.lastSentResultIndex = 0
h.scannedItemsMap = make(map[madmin.HealItemType]int64)
h.healedItemsMap = make(map[madmin.HealItemType]int64)
h.healFailedItemsMap = make(map[string]int64)
}
// getScannedItemsCount - returns a count of all scanned items
func (h *healSequence) getScannedItemsCount() int64 {
var count int64
h.mutex.RLock()
defer h.mutex.RUnlock()
for _, v := range h.scannedItemsMap {
count = count + v
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
return count
}
// getScannedItemsMap - returns map of all scanned items against type
func (h *healSequence) getScannedItemsMap() map[madmin.HealItemType]int64 {
h.mutex.RLock()
defer h.mutex.RUnlock()
// Make a copy before returning the value
retMap := make(map[madmin.HealItemType]int64, len(h.scannedItemsMap))
for k, v := range h.scannedItemsMap {
retMap[k] = v
}
return retMap
}
// getHealedItemsMap - returns the map of all healed items against type
func (h *healSequence) getHealedItemsMap() map[madmin.HealItemType]int64 {
h.mutex.RLock()
defer h.mutex.RUnlock()
// Make a copy before returning the value
retMap := make(map[madmin.HealItemType]int64, len(h.healedItemsMap))
for k, v := range h.healedItemsMap {
retMap[k] = v
}
return retMap
}
// gethealFailedItemsMap - returns map of all items where heal failed against
// drive endpoint and status
func (h *healSequence) gethealFailedItemsMap() map[string]int64 {
h.mutex.RLock()
defer h.mutex.RUnlock()
// Make a copy before returning the value
retMap := make(map[string]int64, len(h.healFailedItemsMap))
for k, v := range h.healFailedItemsMap {
retMap[k] = v
}
return retMap
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
// isQuitting - determines if the heal sequence is quitting (due to an
// external signal)
func (h *healSequence) isQuitting() bool {
select {
case <-h.ctx.Done():
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
return true
default:
return false
}
}
// check if the heal sequence has ended
func (h *healSequence) hasEnded() bool {
h.mutex.RLock()
ended := len(h.currentStatus.Items) == 0 || h.currentStatus.Summary == healStoppedStatus || h.currentStatus.Summary == healFinishedStatus
h.mutex.RUnlock()
return ended
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
// stops the heal sequence - safe to call multiple times.
func (h *healSequence) stop() {
h.cancelCtx()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
// pushHealResultItem - pushes a heal result item for consumption in
// the heal-status API. It blocks if there are
// maxUnconsumedHealResultItems. When it blocks, the heal sequence
// routine is effectively paused - this happens when the server has
// accumulated the maximum number of heal records per heal
// sequence. When the client consumes further records, the heal
// sequence automatically resumes. The return value indicates if the
// operation succeeded.
func (h *healSequence) pushHealResultItem(r madmin.HealResultItem) error {
// start a timer to keep an upper time limit to find an empty
// slot to add the given heal result - if no slot is found it
// means that the server is holding the maximum amount of
// heal-results in memory and the client has not consumed it
// for too long.
unconsumedTimer := time.NewTimer(healUnconsumedTimeout)
defer func() {
// stop the timeout timer so it is garbage collected.
if !unconsumedTimer.Stop() {
<-unconsumedTimer.C
}
}()
var itemsLen int
for {
h.mutex.Lock()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
itemsLen = len(h.currentStatus.Items)
if itemsLen == maxUnconsumedHealResultItems {
// wait for a second, or quit if an external
// stop signal is received or the
// unconsumedTimer fires.
select {
// Check after a second
case <-time.After(time.Second):
h.mutex.Unlock()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
continue
case <-h.ctx.Done():
h.mutex.Unlock()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// discard result and return.
return errHealStopSignalled
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// Timeout if no results consumed for too long.
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
case <-unconsumedTimer.C:
h.mutex.Unlock()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
return errHealIdleTimeout
}
}
break
}
// Set the correct result index for the new result item
if itemsLen > 0 {
r.ResultIndex = 1 + h.currentStatus.Items[itemsLen-1].ResultIndex
} else {
r.ResultIndex = 1 + h.lastSentResultIndex
}
// append to results
h.currentStatus.Items = append(h.currentStatus.Items, r)
// release lock
h.mutex.Unlock()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
return nil
}
// healSequenceStart - this is the top-level background heal
// routine. It launches another go-routine that actually traverses
// on-disk data, checks and heals according to the selected
// settings. This go-routine itself, (1) monitors the traversal
// routine for completion, and (2) listens for external stop
// signals. When either event happens, it sets the finish status for
// the heal-sequence.
func (h *healSequence) healSequenceStart() {
// Set status as running
h.mutex.Lock()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
h.currentStatus.Summary = healRunningStatus
h.currentStatus.StartTime = UTCNow()
h.mutex.Unlock()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
if h.sourceCh == nil {
go h.traverseAndHeal()
} else {
go h.healFromSourceCh()
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
select {
case err, ok := <-h.traverseAndHealDoneCh:
if !ok {
return
}
h.mutex.Lock()
h.endTime = UTCNow()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// Heal traversal is complete.
if err == nil {
// heal traversal succeeded.
h.currentStatus.Summary = healFinishedStatus
} else {
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// heal traversal had an error.
h.currentStatus.Summary = healStoppedStatus
h.currentStatus.FailureDetail = err.Error()
}
h.mutex.Unlock()
case <-h.ctx.Done():
h.mutex.Lock()
h.endTime = UTCNow()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
h.currentStatus.Summary = healStoppedStatus
h.currentStatus.FailureDetail = errHealStopSignalled.Error()
h.mutex.Unlock()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// drain traverse channel so the traversal
// go-routine does not leak.
go func() {
// Eventually the traversal go-routine closes
// the channel and returns, so this go-routine
// itself will not leak.
<-h.traverseAndHealDoneCh
}()
}
}
func (h *healSequence) queueHealTask(source healSource, healType madmin.HealItemType) error {
// Send heal request
task := healTask{
path: source.path,
opts: h.settings,
responseCh: h.respCh,
}
if !source.opts.Equal(h.settings) {
task.opts = source.opts
}
globalBackgroundHealRoutine.queueHealTask(task)
select {
case res := <-h.respCh:
if !h.reportProgress {
// Object might have been deleted, by the time heal
// was attempted, we should ignore this object and
// return success.
if isErrObjectNotFound(res.err) {
return nil
}
h.mutex.Lock()
defer h.mutex.Unlock()
// Progress is not reported in case of background heal processing.
// Instead we increment relevant counter based on the heal result
// for prometheus reporting.
if res.err != nil {
for _, d := range res.result.After.Drives {
// For failed items we report the endpoint and drive state
// This will help users take corrective actions for drives
h.healFailedItemsMap[d.Endpoint+","+d.State]++
}
} else {
// Only object type reported for successful healing
h.healedItemsMap[res.result.Type]++
}
// Report caller of any failure
return res.err
}
res.result.Type = healType
if res.err != nil {
// Object might have been deleted, by the time heal
// was attempted, we should ignore this object and return success.
if isErrObjectNotFound(res.err) {
return nil
}
// Only report object error
if healType != madmin.HealItemObject {
return res.err
}
res.result.Detail = res.err.Error()
}
return h.pushHealResultItem(res.result)
case <-h.ctx.Done():
return nil
}
}
func (h *healSequence) healItemsFromSourceCh() error {
bucketsOnly := true // heal buckets only, not objects.
if err := h.healItems(bucketsOnly); err != nil {
logger.LogIf(h.ctx, err)
}
for {
select {
case source, ok := <-h.sourceCh:
if !ok {
return nil
}
var itemType madmin.HealItemType
switch {
case source.path == nopHeal:
continue
case source.path == SlashSeparator:
itemType = madmin.HealItemMetadata
case !strings.Contains(source.path, SlashSeparator):
itemType = madmin.HealItemBucket
default:
itemType = madmin.HealItemObject
}
if err := h.queueHealTask(source, itemType); err != nil {
logger.LogIf(h.ctx, err)
}
h.scannedItemsMap[itemType]++
h.lastHealActivity = UTCNow()
case <-h.ctx.Done():
return nil
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
}
}
func (h *healSequence) healFromSourceCh() {
h.healItemsFromSourceCh()
}
func (h *healSequence) healItems(bucketsOnly bool) error {
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// Start with format healing
if err := h.healDiskFormat(); err != nil {
return err
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// Start healing the config prefix.
if err := h.healMinioSysMeta(minioConfigPrefix)(); err != nil {
return err
}
// Start healing the bucket config prefix.
if err := h.healMinioSysMeta(bucketConfigPrefix)(); err != nil {
return err
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// Heal buckets and objects
return h.healBuckets(bucketsOnly)
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// traverseAndHeal - traverses on-disk data and performs healing
// according to settings. At each "safe" point it also checks if an
// external quit signal has been received and quits if so. Since the
// healing traversal may be mutating on-disk data when an external
// quit signal is received, this routine cannot quit immediately and
// has to wait until a safe point is reached, such as between scanning
// two objects.
func (h *healSequence) traverseAndHeal() {
bucketsOnly := false // Heals buckets and objects also.
h.traverseAndHealDoneCh <- h.healItems(bucketsOnly)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
close(h.traverseAndHealDoneCh)
}
// healMinioSysMeta - heals all files under a given meta prefix, returns a function
// which in-turn heals the respective meta directory path and any files in int.
func (h *healSequence) healMinioSysMeta(metaPrefix string) func() error {
return func() error {
// Get current object layer instance.
objectAPI := newObjectLayerWithoutSafeModeFn()
if objectAPI == nil {
return errServerNotInitialized
}
// NOTE: Healing on meta is run regardless
// of any bucket being selected, this is to ensure that
// meta are always upto date and correct.
return objectAPI.HealObjects(h.ctx, minioMetaBucket, metaPrefix, h.settings, func(bucket string, object string) error {
if h.isQuitting() {
return errHealStopSignalled
}
herr := h.queueHealTask(healSource{path: pathJoin(bucket, object)}, madmin.HealItemBucketMetadata)
// Object might have been deleted, by the time heal
// was attempted we ignore this object an move on.
if isErrObjectNotFound(herr) {
return nil
}
return herr
})
}
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// healDiskFormat - heals format.json, return value indicates if a
// failure error occurred.
func (h *healSequence) healDiskFormat() error {
if h.isQuitting() {
return errHealStopSignalled
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// Get current object layer instance.
objectAPI := newObjectLayerWithoutSafeModeFn()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
if objectAPI == nil {
return errServerNotInitialized
}
return h.queueHealTask(healSource{path: SlashSeparator}, madmin.HealItemMetadata)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
// healBuckets - check for all buckets heal or just particular bucket.
func (h *healSequence) healBuckets(bucketsOnly bool) error {
if h.isQuitting() {
return errHealStopSignalled
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// 1. If a bucket was specified, heal only the bucket.
if h.bucket != "" {
return h.healBucket(h.bucket, bucketsOnly)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
// Get current object layer instance.
objectAPI := newObjectLayerWithoutSafeModeFn()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
if objectAPI == nil {
return errServerNotInitialized
}
buckets, err := objectAPI.ListBucketsHeal(h.ctx)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
if err != nil {
return errFnHealFromAPIErr(h.ctx, err)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
for _, bucket := range buckets {
if err = h.healBucket(bucket.Name, bucketsOnly); err != nil {
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
return err
}
}
return nil
}
// healBucket - traverses and heals given bucket
func (h *healSequence) healBucket(bucket string, bucketsOnly bool) error {
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
// Get current object layer instance.
objectAPI := newObjectLayerWithoutSafeModeFn()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
if objectAPI == nil {
return errServerNotInitialized
}
if err := h.queueHealTask(healSource{path: bucket}, madmin.HealItemBucket); err != nil {
return err
}
if bucketsOnly {
return nil
}
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
if !h.settings.Recursive {
if h.objPrefix != "" {
// Check if an object named as the objPrefix exists,
// and if so heal it.
_, err := objectAPI.GetObjectInfo(h.ctx, bucket, h.objPrefix, ObjectOptions{})
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
if err == nil {
if err = h.healObject(bucket, h.objPrefix); err != nil {
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
return err
}
}
}
return nil
}
if err := objectAPI.HealObjects(h.ctx, bucket, h.objPrefix, h.settings, h.healObject); err != nil {
return errFnHealFromAPIErr(h.ctx, err)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}
return nil
}
// healObject - heal the given object and record result
func (h *healSequence) healObject(bucket, object string) error {
// Get current object layer instance.
objectAPI := newObjectLayerWithoutSafeModeFn()
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
if objectAPI == nil {
return errServerNotInitialized
}
if h.isQuitting() {
return errHealStopSignalled
}
return h.queueHealTask(healSource{path: pathJoin(bucket, object)}, madmin.HealItemObject)
Move admin APIs to new path and add redesigned heal APIs (#5351) - Changes related to moving admin APIs - admin APIs now have an endpoint under /minio/admin - admin APIs are now versioned - a new API to server the version is added at "GET /minio/admin/version" and all API operations have the path prefix /minio/admin/v1/<operation> - new service stop API added - credentials change API is moved to /minio/admin/v1/config/credential - credentials change API and configuration get/set API now require TLS so that credentials are protected - all API requests now receive JSON - heal APIs are disabled as they will be changed substantially - Heal API changes Heal API is now provided at a single endpoint with the ability for a client to start a heal sequence on all the data in the server, a single bucket, or under a prefix within a bucket. When a heal sequence is started, the server returns a unique token that needs to be used for subsequent 'status' requests to fetch heal results. On each status request from the client, the server returns heal result records that it has accumulated since the previous status request. The server accumulates upto 1000 records and pauses healing further objects until the client requests for status. If the client does not request any further records for a long time, the server aborts the heal sequence automatically. A heal result record is returned for each entity healed on the server, such as system metadata, object metadata, buckets and objects, and has information about the before and after states on each disk. A client may request to force restart a heal sequence - this causes the running heal sequence to be aborted at the next safe spot and starts a new heal sequence.
7 years ago
}