fix: refactor locks to apply them uniquely per node (#11052)

This refactor is done for few reasons below

- to avoid deadlocks in scenarios when number
  of nodes are smaller < actual erasure stripe
  count where in N participating local lockers
  can lead to deadlocks across systems.

- avoids expiry routines to run 1000 of separate
  network operations and routes per disk where
  as each of them are still accessing one single
  local entity.

- it is ideal to have since globalLockServer
  per instance.

- In a 32node deployment however, each server
  group is still concentrated towards the
  same set of lockers that partipicate during
  the write/read phase, unlike previous minio/dsync
  implementation - this potentially avoids send
  32 requests instead we will still send at max
  requests of unique nodes participating in a
  write/read phase.

- reduces overall chattiness on smaller setups.
master
Harshavardhana 4 years ago committed by GitHub
parent 97856bfebf
commit 4550ac6fff
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
  1. 2
      Makefile
  2. 26
      buildscripts/verify-build.sh
  3. 6
      cmd/admin-handlers.go
  4. 16
      cmd/endpoint.go
  5. 23
      cmd/erasure-multipart.go
  6. 66
      cmd/erasure-sets.go
  7. 8
      cmd/local-locker.go
  8. 13
      cmd/lock-rest-client.go
  9. 49
      cmd/lock-rest-server.go
  10. 2
      cmd/namespace-lock.go
  11. 8
      cmd/notification.go
  12. 10
      cmd/peer-rest-client.go
  13. 7
      cmd/peer-rest-server.go
  14. 2
      cmd/routers.go
  15. 12
      pkg/argon2/argon2.go
  16. 9
      pkg/dsync/drwmutex.go
  17. 6
      pkg/madmin/encrypt.go

@ -57,7 +57,7 @@ test-race: verifiers build
# Verify minio binary # Verify minio binary
verify: verify:
@echo "Verifying build with race" @echo "Verifying build with race"
@GO111MODULE=on CGO_ENABLED=1 go build -race -tags kqueue -trimpath --ldflags "$(LDFLAGS)" -o $(PWD)/minio 1>/dev/null @GO111MODULE=on CGO_ENABLED=1 go build -tags kqueue -trimpath --ldflags "$(LDFLAGS)" -o $(PWD)/minio 1>/dev/null
@(env bash $(PWD)/buildscripts/verify-build.sh) @(env bash $(PWD)/buildscripts/verify-build.sh)
# Verify healing of disks with minio binary # Verify healing of disks with minio binary

@ -56,7 +56,8 @@ function start_minio_erasure()
function start_minio_erasure_sets() function start_minio_erasure_sets()
{ {
"${MINIO[@]}" server "${WORK_DIR}/erasure-disk-sets{1...32}" >"$WORK_DIR/erasure-minio-sets.log" 2>&1 & export MINIO_ENDPOINTS="${WORK_DIR}/erasure-disk-sets{1...32}"
"${MINIO[@]}" server > "$WORK_DIR/erasure-minio-sets.log" 2>&1 &
sleep 15 sleep 15
} }
@ -64,9 +65,9 @@ function start_minio_pool_erasure_sets()
{ {
export MINIO_ACCESS_KEY=$ACCESS_KEY export MINIO_ACCESS_KEY=$ACCESS_KEY
export MINIO_SECRET_KEY=$SECRET_KEY export MINIO_SECRET_KEY=$SECRET_KEY
export MINIO_ENDPOINTS="http://127.0.0.1:9000${WORK_DIR}/pool-disk-sets{1...4} http://127.0.0.1:9001${WORK_DIR}/pool-disk-sets{5...8}"
"${MINIO[@]}" server --address=:9000 "http://127.0.0.1:9000${WORK_DIR}/pool-disk-sets{1...4}" "http://127.0.0.1:9001${WORK_DIR}/pool-disk-sets{5...8}" >"$WORK_DIR/pool-minio-9000.log" 2>&1 & "${MINIO[@]}" server --address ":9000" > "$WORK_DIR/pool-minio-9000.log" 2>&1 &
"${MINIO[@]}" server --address=:9001 "http://127.0.0.1:9000${WORK_DIR}/pool-disk-sets{1...4}" "http://127.0.0.1:9001${WORK_DIR}/pool-disk-sets{5...8}" >"$WORK_DIR/pool-minio-9001.log" 2>&1 & "${MINIO[@]}" server --address ":9001" > "$WORK_DIR/pool-minio-9001.log" 2>&1 &
sleep 40 sleep 40
} }
@ -75,9 +76,9 @@ function start_minio_pool_erasure_sets_ipv6()
{ {
export MINIO_ACCESS_KEY=$ACCESS_KEY export MINIO_ACCESS_KEY=$ACCESS_KEY
export MINIO_SECRET_KEY=$SECRET_KEY export MINIO_SECRET_KEY=$SECRET_KEY
export MINIO_ENDPOINTS="http://[::1]:9000${WORK_DIR}/pool-disk-sets{1...4} http://[::1]:9001${WORK_DIR}/pool-disk-sets{5...8}"
"${MINIO[@]}" server --address="[::1]:9000" "http://[::1]:9000${WORK_DIR}/pool-disk-sets{1...4}" "http://[::1]:9001${WORK_DIR}/pool-disk-sets{5...8}" >"$WORK_DIR/pool-minio-ipv6-9000.log" 2>&1 & "${MINIO[@]}" server --address="[::1]:9000" > "$WORK_DIR/pool-minio-ipv6-9000.log" 2>&1 &
"${MINIO[@]}" server --address="[::1]:9001" "http://[::1]:9000${WORK_DIR}/pool-disk-sets{1...4}" "http://[::1]:9001${WORK_DIR}/pool-disk-sets{5...8}" >"$WORK_DIR/pool-minio-ipv6-9001.log" 2>&1 & "${MINIO[@]}" server --address="[::1]:9001" > "$WORK_DIR/pool-minio-ipv6-9001.log" 2>&1 &
sleep 40 sleep 40
} }
@ -86,10 +87,10 @@ function start_minio_dist_erasure()
{ {
export MINIO_ACCESS_KEY=$ACCESS_KEY export MINIO_ACCESS_KEY=$ACCESS_KEY
export MINIO_SECRET_KEY=$SECRET_KEY export MINIO_SECRET_KEY=$SECRET_KEY
"${MINIO[@]}" server --address=:9000 "http://127.0.0.1:9000${WORK_DIR}/dist-disk1" "http://127.0.0.1:9001${WORK_DIR}/dist-disk2" "http://127.0.0.1:9002${WORK_DIR}/dist-disk3" "http://127.0.0.1:9003${WORK_DIR}/dist-disk4" >"$WORK_DIR/dist-minio-9000.log" 2>&1 & export MINIO_ENDPOINTS="http://127.0.0.1:9000${WORK_DIR}/dist-disk1 http://127.0.0.1:9001${WORK_DIR}/dist-disk2 http://127.0.0.1:9002${WORK_DIR}/dist-disk3 http://127.0.0.1:9003${WORK_DIR}/dist-disk4"
"${MINIO[@]}" server --address=:9001 "http://127.0.0.1:9000${WORK_DIR}/dist-disk1" "http://127.0.0.1:9001${WORK_DIR}/dist-disk2" "http://127.0.0.1:9002${WORK_DIR}/dist-disk3" "http://127.0.0.1:9003${WORK_DIR}/dist-disk4" >"$WORK_DIR/dist-minio-9001.log" 2>&1 & for i in $(seq 0 3); do
"${MINIO[@]}" server --address=:9002 "http://127.0.0.1:9000${WORK_DIR}/dist-disk1" "http://127.0.0.1:9001${WORK_DIR}/dist-disk2" "http://127.0.0.1:9002${WORK_DIR}/dist-disk3" "http://127.0.0.1:9003${WORK_DIR}/dist-disk4" >"$WORK_DIR/dist-minio-9002.log" 2>&1 & "${MINIO[@]}" server --address ":900${i}" > "$WORK_DIR/dist-minio-900${i}.log" 2>&1 &
"${MINIO[@]}" server --address=:9003 "http://127.0.0.1:9000${WORK_DIR}/dist-disk1" "http://127.0.0.1:9001${WORK_DIR}/dist-disk2" "http://127.0.0.1:9002${WORK_DIR}/dist-disk3" "http://127.0.0.1:9003${WORK_DIR}/dist-disk4" >"$WORK_DIR/dist-minio-9003.log" 2>&1 & done
sleep 40 sleep 40
} }
@ -112,7 +113,8 @@ function run_test_fs()
return "$rv" return "$rv"
} }
function run_test_erasure_sets() { function run_test_erasure_sets()
{
start_minio_erasure_sets start_minio_erasure_sets
(cd "$WORK_DIR" && "$FUNCTIONAL_TESTS") (cd "$WORK_DIR" && "$FUNCTIONAL_TESTS")

@ -373,8 +373,7 @@ func topLockEntries(peerLocks []*PeerLocks, stale bool) madmin.LockEntries {
if peerLock == nil { if peerLock == nil {
continue continue
} }
for _, locks := range peerLock.Locks { for k, v := range peerLock.Locks {
for k, v := range locks {
for _, lockReqInfo := range v { for _, lockReqInfo := range v {
if val, ok := entryMap[lockReqInfo.UID]; ok { if val, ok := entryMap[lockReqInfo.UID]; ok {
val.ServerList = append(val.ServerList, peerLock.Addr) val.ServerList = append(val.ServerList, peerLock.Addr)
@ -384,7 +383,6 @@ func topLockEntries(peerLocks []*PeerLocks, stale bool) madmin.LockEntries {
} }
} }
} }
}
var lockEntries madmin.LockEntries var lockEntries madmin.LockEntries
for _, v := range entryMap { for _, v := range entryMap {
if stale { if stale {
@ -402,7 +400,7 @@ func topLockEntries(peerLocks []*PeerLocks, stale bool) madmin.LockEntries {
// PeerLocks holds server information result of one node // PeerLocks holds server information result of one node
type PeerLocks struct { type PeerLocks struct {
Addr string Addr string
Locks GetLocksResp Locks map[string][]lockRequesterInfo
} }
// TopLocksHandler Get list of locks in use // TopLocksHandler Get list of locks in use

@ -242,6 +242,22 @@ func (l *EndpointServerPools) Add(zeps ZoneEndpoints) error {
return nil return nil
} }
// Localhost - returns the local hostname from list of endpoints
func (l EndpointServerPools) Localhost() string {
for _, ep := range l {
for _, endpoint := range ep.Endpoints {
if endpoint.IsLocal {
u := &url.URL{
Scheme: endpoint.Scheme,
Host: endpoint.Host,
}
return u.String()
}
}
}
return ""
}
// FirstLocal returns true if the first endpoint is local. // FirstLocal returns true if the first endpoint is local.
func (l EndpointServerPools) FirstLocal() bool { func (l EndpointServerPools) FirstLocal() bool {
return l[0].Endpoints[0].IsLocal return l[0].Endpoints[0].IsLocal

@ -89,27 +89,12 @@ func (er erasureObjects) removeObjectPart(bucket, object, uploadID, dataDir stri
} }
// Clean-up the old multipart uploads. Should be run in a Go routine. // Clean-up the old multipart uploads. Should be run in a Go routine.
func (er erasureObjects) cleanupStaleUploads(ctx context.Context, cleanupInterval, expiry time.Duration) { func (er erasureObjects) cleanupStaleUploads(ctx context.Context, expiry time.Duration) {
ticker := time.NewTicker(cleanupInterval)
defer ticker.Stop()
for {
select {
case <-ctx.Done():
return
case <-ticker.C:
var disk StorageAPI
// run multiple cleanup's local to this server. // run multiple cleanup's local to this server.
for _, d := range er.getLoadBalancedLocalDisks() { for _, disk := range er.getLoadBalancedLocalDisks() {
if d != nil { if disk != nil {
disk = d
break
}
}
if disk == nil {
continue
}
er.cleanupStaleUploadsOnDisk(ctx, disk, expiry) er.cleanupStaleUploadsOnDisk(ctx, disk, expiry)
return
} }
} }
} }

@ -30,6 +30,7 @@ import (
"github.com/dchest/siphash" "github.com/dchest/siphash"
"github.com/google/uuid" "github.com/google/uuid"
"github.com/minio/minio-go/v7/pkg/set"
"github.com/minio/minio-go/v7/pkg/tags" "github.com/minio/minio-go/v7/pkg/tags"
"github.com/minio/minio/cmd/logger" "github.com/minio/minio/cmd/logger"
"github.com/minio/minio/pkg/bpool" "github.com/minio/minio/pkg/bpool"
@ -278,10 +279,20 @@ func (s *erasureSets) monitorAndConnectEndpoints(ctx context.Context, monitorInt
// GetAllLockers return a list of all lockers for all sets. // GetAllLockers return a list of all lockers for all sets.
func (s *erasureSets) GetAllLockers() []dsync.NetLocker { func (s *erasureSets) GetAllLockers() []dsync.NetLocker {
allLockers := make([]dsync.NetLocker, s.setDriveCount*s.setCount) var allLockers []dsync.NetLocker
for i, lockers := range s.erasureLockers { lockEpSet := set.NewStringSet()
for j, locker := range lockers { for _, lockers := range s.erasureLockers {
allLockers[i*s.setDriveCount+j] = locker for _, locker := range lockers {
if locker == nil || !locker.IsOnline() {
// Skip any offline lockers.
continue
}
if lockEpSet.Contains(locker.String()) {
// Skip duplicate lockers.
continue
}
lockEpSet.Add(locker.String())
allLockers = append(allLockers, locker)
} }
} }
return allLockers return allLockers
@ -289,15 +300,8 @@ func (s *erasureSets) GetAllLockers() []dsync.NetLocker {
func (s *erasureSets) GetLockers(setIndex int) func() ([]dsync.NetLocker, string) { func (s *erasureSets) GetLockers(setIndex int) func() ([]dsync.NetLocker, string) {
return func() ([]dsync.NetLocker, string) { return func() ([]dsync.NetLocker, string) {
lockers := make([]dsync.NetLocker, s.setDriveCount) lockers := make([]dsync.NetLocker, len(s.erasureLockers[setIndex]))
copy(lockers, s.erasureLockers[setIndex]) copy(lockers, s.erasureLockers[setIndex])
sort.Slice(lockers, func(i, j int) bool {
// re-order lockers with affinity for
// - non-local lockers
// - online lockers
// are used first
return !lockers[i].IsLocal() && lockers[i].IsOnline()
})
return lockers, s.erasureLockOwner return lockers, s.erasureLockOwner
} }
} }
@ -362,14 +366,24 @@ func newErasureSets(ctx context.Context, endpoints Endpoints, storageDisks []Sto
for i := 0; i < setCount; i++ { for i := 0; i < setCount; i++ {
s.erasureDisks[i] = make([]StorageAPI, setDriveCount) s.erasureDisks[i] = make([]StorageAPI, setDriveCount)
s.erasureLockers[i] = make([]dsync.NetLocker, setDriveCount) }
var erasureLockers = map[string]dsync.NetLocker{}
for _, endpoint := range endpoints {
if _, ok := erasureLockers[endpoint.Host]; !ok {
erasureLockers[endpoint.Host] = newLockAPI(endpoint)
}
} }
for i := 0; i < setCount; i++ { for i := 0; i < setCount; i++ {
var lockerEpSet = set.NewStringSet()
for j := 0; j < setDriveCount; j++ { for j := 0; j < setDriveCount; j++ {
endpoint := endpoints[i*setDriveCount+j] endpoint := endpoints[i*setDriveCount+j]
// Rely on endpoints list to initialize, init lockers and available disks. // Only add lockers only one per endpoint and per erasure set.
s.erasureLockers[i][j] = newLockAPI(endpoint) if locker, ok := erasureLockers[endpoint.Host]; ok && !lockerEpSet.Contains(endpoint.Host) {
lockerEpSet.Add(endpoint.Host)
s.erasureLockers[i] = append(s.erasureLockers[i], locker)
}
disk := storageDisks[i*setDriveCount+j] disk := storageDisks[i*setDriveCount+j]
if disk == nil { if disk == nil {
continue continue
@ -396,11 +410,11 @@ func newErasureSets(ctx context.Context, endpoints Endpoints, storageDisks []Sto
bp: bp, bp: bp,
mrfOpCh: make(chan partialOperation, 10000), mrfOpCh: make(chan partialOperation, 10000),
} }
go s.sets[i].cleanupStaleUploads(ctx,
GlobalStaleUploadsCleanupInterval, GlobalStaleUploadsExpiry)
} }
// start cleanup stale uploads go-routine.
go s.cleanupStaleUploads(ctx, GlobalStaleUploadsCleanupInterval, GlobalStaleUploadsExpiry)
// Start the disk monitoring and connect routine. // Start the disk monitoring and connect routine.
go s.monitorAndConnectEndpoints(ctx, defaultMonitorConnectEndpointInterval) go s.monitorAndConnectEndpoints(ctx, defaultMonitorConnectEndpointInterval)
go s.maintainMRFList() go s.maintainMRFList()
@ -409,6 +423,22 @@ func newErasureSets(ctx context.Context, endpoints Endpoints, storageDisks []Sto
return s, nil return s, nil
} }
func (s *erasureSets) cleanupStaleUploads(ctx context.Context, cleanupInterval, expiry time.Duration) {
ticker := time.NewTicker(cleanupInterval)
defer ticker.Stop()
for {
select {
case <-ctx.Done():
return
case <-ticker.C:
for _, set := range s.sets {
set.cleanupStaleUploads(ctx, expiry)
}
}
}
}
// NewNSLock - initialize a new namespace RWLocker instance. // NewNSLock - initialize a new namespace RWLocker instance.
func (s *erasureSets) NewNSLock(bucket string, objects ...string) RWLocker { func (s *erasureSets) NewNSLock(bucket string, objects ...string) RWLocker {
if len(objects) == 1 { if len(objects) == 1 {

@ -47,12 +47,11 @@ func isWriteLock(lri []lockRequesterInfo) bool {
// localLocker implements Dsync.NetLocker // localLocker implements Dsync.NetLocker
type localLocker struct { type localLocker struct {
mutex sync.Mutex mutex sync.Mutex
endpoint Endpoint
lockMap map[string][]lockRequesterInfo lockMap map[string][]lockRequesterInfo
} }
func (l *localLocker) String() string { func (l *localLocker) String() string {
return l.endpoint.String() return globalEndpoints.Localhost()
} }
func (l *localLocker) canTakeUnlock(resources ...string) bool { func (l *localLocker) canTakeUnlock(resources ...string) bool {
@ -194,7 +193,7 @@ func (l *localLocker) DupLockMap() map[string][]lockRequesterInfo {
l.mutex.Lock() l.mutex.Lock()
defer l.mutex.Unlock() defer l.mutex.Unlock()
lockCopy := make(map[string][]lockRequesterInfo) lockCopy := map[string][]lockRequesterInfo{}
for k, v := range l.lockMap { for k, v := range l.lockMap {
lockCopy[k] = append(lockCopy[k], v...) lockCopy[k] = append(lockCopy[k], v...)
} }
@ -253,9 +252,8 @@ func (l *localLocker) removeEntryIfExists(nlrip nameLockRequesterInfoPair) {
} }
} }
func newLocker(endpoint Endpoint) *localLocker { func newLocker() *localLocker {
return &localLocker{ return &localLocker{
endpoint: endpoint,
lockMap: make(map[string][]lockRequesterInfo), lockMap: make(map[string][]lockRequesterInfo),
} }
} }

@ -32,7 +32,7 @@ import (
// lockRESTClient is authenticable lock REST client // lockRESTClient is authenticable lock REST client
type lockRESTClient struct { type lockRESTClient struct {
restClient *rest.Client restClient *rest.Client
endpoint Endpoint u *url.URL
} }
func toLockError(err error) error { func toLockError(err error) error {
@ -51,7 +51,7 @@ func toLockError(err error) error {
// String stringer *dsync.NetLocker* interface compatible method. // String stringer *dsync.NetLocker* interface compatible method.
func (client *lockRESTClient) String() string { func (client *lockRESTClient) String() string {
return client.endpoint.String() return client.u.String()
} }
// Wrapper to restClient.Call to handle network errors, in case of network error the connection is marked disconnected // Wrapper to restClient.Call to handle network errors, in case of network error the connection is marked disconnected
@ -137,7 +137,7 @@ func (client *lockRESTClient) Expired(ctx context.Context, args dsync.LockArgs)
func newLockAPI(endpoint Endpoint) dsync.NetLocker { func newLockAPI(endpoint Endpoint) dsync.NetLocker {
if endpoint.IsLocal { if endpoint.IsLocal {
return globalLockServers[endpoint] return globalLockServer
} }
return newlockRESTClient(endpoint) return newlockRESTClient(endpoint)
} }
@ -147,7 +147,7 @@ func newlockRESTClient(endpoint Endpoint) *lockRESTClient {
serverURL := &url.URL{ serverURL := &url.URL{
Scheme: endpoint.Scheme, Scheme: endpoint.Scheme,
Host: endpoint.Host, Host: endpoint.Host,
Path: pathJoin(lockRESTPrefix, endpoint.Path, lockRESTVersion), Path: pathJoin(lockRESTPrefix, lockRESTVersion),
} }
restClient := rest.NewClient(serverURL, globalInternodeTransport, newAuthToken) restClient := rest.NewClient(serverURL, globalInternodeTransport, newAuthToken)
@ -163,5 +163,8 @@ func newlockRESTClient(endpoint Endpoint) *lockRESTClient {
return !isNetworkError(err) return !isNetworkError(err)
} }
return &lockRESTClient{endpoint: endpoint, restClient: restClient} return &lockRESTClient{u: &url.URL{
Scheme: endpoint.Scheme,
Host: endpoint.Host,
}, restClient: restClient}
} }

@ -22,7 +22,6 @@ import (
"errors" "errors"
"math/rand" "math/rand"
"net/http" "net/http"
"path"
"sort" "sort"
"strconv" "strconv"
"time" "time"
@ -217,16 +216,14 @@ type nameLockRequesterInfoPair struct {
// getLongLivedLocks returns locks that are older than a certain time and // getLongLivedLocks returns locks that are older than a certain time and
// have not been 'checked' for validity too soon enough // have not been 'checked' for validity too soon enough
func getLongLivedLocks(interval time.Duration) map[Endpoint][]nameLockRequesterInfoPair { func getLongLivedLocks(interval time.Duration) []nameLockRequesterInfoPair {
nlripMap := make(map[Endpoint][]nameLockRequesterInfoPair) nlrip := []nameLockRequesterInfoPair{}
for endpoint, locker := range globalLockServers { globalLockServer.mutex.Lock()
rslt := []nameLockRequesterInfoPair{} for name, lriArray := range globalLockServer.lockMap {
locker.mutex.Lock()
for name, lriArray := range locker.lockMap {
for idx := range lriArray { for idx := range lriArray {
// Check whether enough time has gone by since last check // Check whether enough time has gone by since last check
if time.Since(lriArray[idx].TimeLastCheck) >= interval { if time.Since(lriArray[idx].TimeLastCheck) >= interval {
rslt = append(rslt, nameLockRequesterInfoPair{ nlrip = append(nlrip, nameLockRequesterInfoPair{
name: name, name: name,
lri: lriArray[idx], lri: lriArray[idx],
}) })
@ -234,10 +231,8 @@ func getLongLivedLocks(interval time.Duration) map[Endpoint][]nameLockRequesterI
} }
} }
} }
nlripMap[endpoint] = rslt globalLockServer.mutex.Unlock()
locker.mutex.Unlock() return nlrip
}
return nlripMap
} }
// lockMaintenance loops over locks that have been active for some time and checks back // lockMaintenance loops over locks that have been active for some time and checks back
@ -277,18 +272,12 @@ func lockMaintenance(ctx context.Context, interval time.Duration) error {
} }
} }
allLockersFn := z.GetAllLockers
// Validate if long lived locks are indeed clean. // Validate if long lived locks are indeed clean.
// Get list of long lived locks to check for staleness. // Get list of long lived locks to check for staleness.
for lendpoint, nlrips := range getLongLivedLocks(interval) { nlrips := getLongLivedLocks(interval)
nlripsMap := make(map[string]nlock, len(nlrips)) nlripsMap := make(map[string]nlock, len(nlrips))
for _, nlrip := range nlrips { for _, nlrip := range nlrips {
for _, c := range allLockersFn() { for _, c := range z.GetAllLockers() {
if !c.IsOnline() || c == nil {
continue
}
ctx, cancel := context.WithTimeout(GlobalContext, 5*time.Second) ctx, cancel := context.WithTimeout(GlobalContext, 5*time.Second)
// Call back to original server verify whether the lock is // Call back to original server verify whether the lock is
@ -312,9 +301,7 @@ func lockMaintenance(ctx context.Context, interval time.Duration) error {
// less than the quorum, we have locks expired. // less than the quorum, we have locks expired.
if nlripsMap[nlrip.name].locks < nlrip.lri.Quorum { if nlripsMap[nlrip.name].locks < nlrip.lri.Quorum {
// Purge the stale entry if it exists. // Purge the stale entry if it exists.
globalLockServers[lendpoint].removeEntryIfExists(nlrip) globalLockServer.removeEntryIfExists(nlrip)
}
} }
} }
@ -361,18 +348,12 @@ func startLockMaintenance(ctx context.Context) {
} }
// registerLockRESTHandlers - register lock rest router. // registerLockRESTHandlers - register lock rest router.
func registerLockRESTHandlers(router *mux.Router, endpointServerPools EndpointServerPools) { func registerLockRESTHandlers(router *mux.Router) {
for _, ep := range endpointServerPools {
for _, endpoint := range ep.Endpoints {
if !endpoint.IsLocal {
continue
}
lockServer := &lockRESTServer{ lockServer := &lockRESTServer{
ll: newLocker(endpoint), ll: newLocker(),
} }
subrouter := router.PathPrefix(path.Join(lockRESTPrefix, endpoint.Path)).Subrouter() subrouter := router.PathPrefix(lockRESTPrefix).Subrouter()
subrouter.Methods(http.MethodPost).Path(lockRESTVersionPrefix + lockRESTMethodHealth).HandlerFunc(httpTraceHdrs(lockServer.HealthHandler)) subrouter.Methods(http.MethodPost).Path(lockRESTVersionPrefix + lockRESTMethodHealth).HandlerFunc(httpTraceHdrs(lockServer.HealthHandler))
subrouter.Methods(http.MethodPost).Path(lockRESTVersionPrefix + lockRESTMethodLock).HandlerFunc(httpTraceHdrs(lockServer.LockHandler)) subrouter.Methods(http.MethodPost).Path(lockRESTVersionPrefix + lockRESTMethodLock).HandlerFunc(httpTraceHdrs(lockServer.LockHandler))
subrouter.Methods(http.MethodPost).Path(lockRESTVersionPrefix + lockRESTMethodRLock).HandlerFunc(httpTraceHdrs(lockServer.RLockHandler)) subrouter.Methods(http.MethodPost).Path(lockRESTVersionPrefix + lockRESTMethodRLock).HandlerFunc(httpTraceHdrs(lockServer.RLockHandler))
@ -380,9 +361,7 @@ func registerLockRESTHandlers(router *mux.Router, endpointServerPools EndpointSe
subrouter.Methods(http.MethodPost).Path(lockRESTVersionPrefix + lockRESTMethodRUnlock).HandlerFunc(httpTraceHdrs(lockServer.RUnlockHandler)) subrouter.Methods(http.MethodPost).Path(lockRESTVersionPrefix + lockRESTMethodRUnlock).HandlerFunc(httpTraceHdrs(lockServer.RUnlockHandler))
subrouter.Methods(http.MethodPost).Path(lockRESTVersionPrefix + lockRESTMethodExpired).HandlerFunc(httpTraceAll(lockServer.ExpiredHandler)) subrouter.Methods(http.MethodPost).Path(lockRESTVersionPrefix + lockRESTMethodExpired).HandlerFunc(httpTraceAll(lockServer.ExpiredHandler))
globalLockServers[endpoint] = lockServer.ll globalLockServer = lockServer.ll
}
}
go startLockMaintenance(GlobalContext) go startLockMaintenance(GlobalContext)
} }

@ -34,7 +34,7 @@ import (
) )
// local lock servers // local lock servers
var globalLockServers = make(map[Endpoint]*localLocker) var globalLockServer *localLocker
// RWLocker - locker interface to introduce GetRLock, RUnlock. // RWLocker - locker interface to introduce GetRLock, RUnlock.
type RWLocker interface { type RWLocker interface {

@ -639,15 +639,9 @@ func (sys *NotificationSys) GetLocks(ctx context.Context, r *http.Request) []*Pe
ctx := logger.SetReqInfo(ctx, reqInfo) ctx := logger.SetReqInfo(ctx, reqInfo)
logger.LogOnceIf(ctx, err, sys.peerClients[index].host.String()) logger.LogOnceIf(ctx, err, sys.peerClients[index].host.String())
} }
// Once we have received all the locks currently used from peers
// add the local peer locks list as well.
llockers := make(GetLocksResp, 0, len(globalLockServers))
for _, llocker := range globalLockServers {
llockers = append(llockers, llocker.DupLockMap())
}
locksResp = append(locksResp, &PeerLocks{ locksResp = append(locksResp, &PeerLocks{
Addr: getHostName(r), Addr: getHostName(r),
Locks: llockers, Locks: globalLockServer.DupLockMap(),
}) })
return locksResp return locksResp
} }

@ -84,18 +84,16 @@ func (client *peerRESTClient) Close() error {
return nil return nil
} }
// GetLocksResp stores various info from the client for each lock that is requested.
type GetLocksResp []map[string][]lockRequesterInfo
// GetLocks - fetch older locks for a remote node. // GetLocks - fetch older locks for a remote node.
func (client *peerRESTClient) GetLocks() (locks GetLocksResp, err error) { func (client *peerRESTClient) GetLocks() (lockMap map[string][]lockRequesterInfo, err error) {
respBody, err := client.call(peerRESTMethodGetLocks, nil, nil, -1) respBody, err := client.call(peerRESTMethodGetLocks, nil, nil, -1)
if err != nil { if err != nil {
return return
} }
lockMap = map[string][]lockRequesterInfo{}
defer http.DrainBody(respBody) defer http.DrainBody(respBody)
err = gob.NewDecoder(respBody).Decode(&locks) err = gob.NewDecoder(respBody).Decode(&lockMap)
return locks, err return lockMap, err
} }
// ServerInfo - fetch server information for a remote node. // ServerInfo - fetch server information for a remote node.

@ -48,12 +48,7 @@ func (s *peerRESTServer) GetLocksHandler(w http.ResponseWriter, r *http.Request)
} }
ctx := newContext(r, w, "GetLocks") ctx := newContext(r, w, "GetLocks")
logger.LogIf(ctx, gob.NewEncoder(w).Encode(globalLockServer.DupLockMap()))
llockers := make(GetLocksResp, 0, len(globalLockServers))
for _, llocker := range globalLockServers {
llockers = append(llockers, llocker.DupLockMap())
}
logger.LogIf(ctx, gob.NewEncoder(w).Encode(llockers))
w.(http.Flusher).Flush() w.(http.Flusher).Flush()

@ -34,7 +34,7 @@ func registerDistErasureRouters(router *mux.Router, endpointServerPools Endpoint
registerBootstrapRESTHandlers(router) registerBootstrapRESTHandlers(router)
// Register distributed namespace lock routers. // Register distributed namespace lock routers.
registerLockRESTHandlers(router, endpointServerPools) registerLockRESTHandlers(router)
} }
// List of some generic handlers which are applied for all incoming requests. // List of some generic handlers which are applied for all incoming requests.

@ -130,21 +130,21 @@ func NewIDKey(time, memory uint32, threads uint8) func([]byte, []byte, []byte, [
pool := sync.Pool{ pool := sync.Pool{
New: func() interface{} { New: func() interface{} {
b := make([]block, memory) b := make([]block, memory)
return b return &b
}, },
} }
return func(password, salt, secret, data []byte, keyLen uint32) []byte { return func(password, salt, secret, data []byte, keyLen uint32) []byte {
B := pool.Get().([]block) B := pool.Get().(*[]block)
defer func() { defer func() {
clearBlocks(B) clearBlocks(*B)
pool.Put(B) pool.Put(B)
}() }()
h0 := initHash(password, salt, secret, data, time, hashMemory, uint32(threads), keyLen, argon2id) h0 := initHash(password, salt, secret, data, time, hashMemory, uint32(threads), keyLen, argon2id)
B = initBlocks(&h0, B, uint32(threads)) B1 := initBlocks(&h0, *B, uint32(threads))
processBlocks(B, time, memory, uint32(threads), argon2id) processBlocks(B1, time, memory, uint32(threads), argon2id)
return extractKey(B, memory, uint32(threads), keyLen) return extractKey(B1, memory, uint32(threads), keyLen)
} }
} }

@ -299,7 +299,9 @@ func lock(ctx context.Context, ds *Dsync, locks *[]string, id, source string, is
quorumLocked := checkQuorumLocked(locks, quorum) && locksFailed <= tolerance quorumLocked := checkQuorumLocked(locks, quorum) && locksFailed <= tolerance
if !quorumLocked { if !quorumLocked {
log("Releasing all acquired locks now abandoned after quorum was not met\n") log("Releasing all acquired locks now abandoned after quorum was not met\n")
releaseAll(ds, tolerance, owner, locks, isReadLock, restClnts, lockNames...) if !releaseAll(ds, tolerance, owner, locks, isReadLock, restClnts, lockNames...) {
log("Unable to release acquired locks, stale locks might be present\n")
}
} }
// We may have some unused results in ch, release them async. // We may have some unused results in ch, release them async.
@ -308,11 +310,10 @@ func lock(ctx context.Context, ds *Dsync, locks *[]string, id, source string, is
close(ch) close(ch)
for grantToBeReleased := range ch { for grantToBeReleased := range ch {
if grantToBeReleased.isLocked() { if grantToBeReleased.isLocked() {
// release lock // release abandoned lock
log("Releasing abandoned lock\n") log("Releasing abandoned lock\n")
sendRelease(ds, restClnts[grantToBeReleased.index], sendRelease(ds, restClnts[grantToBeReleased.index],
owner, owner, grantToBeReleased.lockUID, isReadLock, lockNames...)
grantToBeReleased.lockUID, isReadLock, lockNames...)
} }
} }
}() }()

@ -28,7 +28,11 @@ import (
"github.com/secure-io/sio-go/sioutil" "github.com/secure-io/sio-go/sioutil"
) )
var idKey = argon2.NewIDKey(1, 64*1024, 4) var idKey func([]byte, []byte, []byte, []byte, uint32) []byte
func init() {
idKey = argon2.NewIDKey(1, 64*1024, 4)
}
// EncryptData encrypts the data with an unique key // EncryptData encrypts the data with an unique key
// derived from password using the Argon2id PBKDF. // derived from password using the Argon2id PBKDF.

Loading…
Cancel
Save