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.

449 lines
12 KiB

//
// Copyright (c) 2018, Joyent, Inc. All rights reserved.
//
// This Source Code Form is subject to the terms of the Mozilla Public
// License, v. 2.0. If a copy of the MPL was not distributed with this
// file, You can obtain one at http://mozilla.org/MPL/2.0/.
//
package storage
import (
"context"
"encoding/json"
"io"
"net/http"
"net/url"
"path"
"strconv"
"strings"
"time"
"github.com/joyent/triton-go/client"
"github.com/pkg/errors"
)
type JobClient struct {
client *client.Client
}
const (
JobStateDone = "done"
JobStateRunning = "running"
)
// JobPhase represents the specification for a map or reduce phase of a Manta
// job.
type JobPhase struct {
// Type is the type of phase. Must be `map` or `reduce`.
Type string `json:"type,omitempty"`
// Assets is an array of objects to be placed in your compute zones.
Assets []string `json:"assets,omitempty"`
// Exec is the shell statement to execute. It may be any valid shell
// command, including pipelines and other shell syntax. You can also
// execute programs stored in the service by including them in "assets"
// and referencing them as /assets/$manta_path.
Exec string `json:"exec"`
// Init is a shell statement to execute in each compute zone before
// any tasks are executed. The same constraints apply as to Exec.
Init string `json:"init"`
// ReducerCount is an optional number of reducers for this phase. The
// default value if not specified is 1. The maximum value is 1024.
ReducerCount uint `json:"count,omitempty"`
// Memory is the amount of DRAM in MB to be allocated to the compute
// zone. Valid values are 256, 512, 1024, 2048, 4096 or 8192.
Memory uint64 `json:"memory,omitempty"`
// Disk is the amount of disk space in GB to be allocated to the compute
// zone. Valid values are 2, 4, 8, 16, 32, 64, 128, 256, 512 or 1024.
Disk uint64 `json:"disk,omitempty"`
}
// JobSummary represents the summary of a compute job in Manta.
type JobSummary struct {
ModifiedTime time.Time `json:"mtime"`
ID string `json:"name"`
}
// Job represents a compute job in Manta.
type Job struct {
ID string `json:"id"`
Name string `json:"name"`
Phases []*JobPhase `json:"phases"`
State string `json:"state"`
Cancelled bool `json:"cancelled"`
InputDone bool `json:"inputDone"`
CreatedTime time.Time `json:"timeCreated"`
DoneTime time.Time `json:"timeDone"`
Transient bool `json:"transient"`
Stats *JobStats `json:"stats"`
}
// JobStats represents statistics for a compute job in Manta.
type JobStats struct {
Errors uint64 `json:"errors"`
Outputs uint64 `json:"outputs"`
Retries uint64 `json:"retries"`
Tasks uint64 `json:"tasks"`
TasksDone uint64 `json:"tasksDone"`
}
// CreateJobInput represents parameters to a CreateJob operation.
type CreateJobInput struct {
Name string `json:"name"`
Phases []*JobPhase `json:"phases"`
}
// CreateJobOutput contains the outputs of a CreateJob operation.
type CreateJobOutput struct {
JobID string
}
// CreateJob submits a new job to be executed. This call is not
// idempotent, so calling it twice will create two jobs.
func (s *JobClient) Create(ctx context.Context, input *CreateJobInput) (*CreateJobOutput, error) {
fullPath := path.Join("/", s.client.AccountName, "jobs")
reqInput := client.RequestInput{
Method: http.MethodPost,
Path: fullPath,
Body: input,
}
respBody, respHeaders, err := s.client.ExecuteRequestStorage(ctx, reqInput)
if respBody != nil {
defer respBody.Close()
}
if err != nil {
return nil, errors.Wrap(err, "unable to create job")
}
jobURI := respHeaders.Get("Location")
parts := strings.Split(jobURI, "/")
jobID := parts[len(parts)-1]
response := &CreateJobOutput{
JobID: jobID,
}
return response, nil
}
// AddJobInputs represents parameters to a AddJobInputs operation.
type AddJobInputsInput struct {
JobID string
ObjectPaths []string
}
// AddJobInputs submits inputs to an already created job.
func (s *JobClient) AddInputs(ctx context.Context, input *AddJobInputsInput) error {
fullPath := path.Join("/", s.client.AccountName, "jobs", input.JobID, "live", "in")
headers := &http.Header{}
headers.Set("Content-Type", "text/plain")
reader := strings.NewReader(strings.Join(input.ObjectPaths, "\n"))
reqInput := client.RequestNoEncodeInput{
Method: http.MethodPost,
Path: fullPath,
Headers: headers,
Body: reader,
}
respBody, _, err := s.client.ExecuteRequestNoEncode(ctx, reqInput)
if respBody != nil {
defer respBody.Close()
}
if err != nil {
return errors.Wrap(err, "unable to add job inputs")
}
return nil
}
// EndJobInputInput represents parameters to a EndJobInput operation.
type EndJobInputInput struct {
JobID string
}
// EndJobInput submits inputs to an already created job.
func (s *JobClient) EndInput(ctx context.Context, input *EndJobInputInput) error {
fullPath := path.Join("/", s.client.AccountName, "jobs", input.JobID, "live", "in", "end")
reqInput := client.RequestNoEncodeInput{
Method: http.MethodPost,
Path: fullPath,
}
respBody, _, err := s.client.ExecuteRequestNoEncode(ctx, reqInput)
if respBody != nil {
defer respBody.Close()
}
if err != nil {
return errors.Wrap(err, "unable to end job inputs")
}
return nil
}
// CancelJobInput represents parameters to a CancelJob operation.
type CancelJobInput struct {
JobID string
}
// CancelJob cancels a job from doing any further work. Cancellation
// is asynchronous and "best effort"; there is no guarantee the job
// will actually stop. For example, short jobs where input is already
// closed will likely still run to completion.
//
// This is however useful when:
// - input is still open
// - you have a long-running job
func (s *JobClient) Cancel(ctx context.Context, input *CancelJobInput) error {
fullPath := path.Join("/", s.client.AccountName, "jobs", input.JobID, "live", "cancel")
reqInput := client.RequestNoEncodeInput{
Method: http.MethodPost,
Path: fullPath,
}
respBody, _, err := s.client.ExecuteRequestNoEncode(ctx, reqInput)
if respBody != nil {
defer respBody.Close()
}
if err != nil {
return errors.Wrap(err, "unable to cancel job")
}
return nil
}
// ListJobsInput represents parameters to a ListJobs operation.
type ListJobsInput struct {
RunningOnly bool
Limit uint64
Marker string
}
// ListJobsOutput contains the outputs of a ListJobs operation.
type ListJobsOutput struct {
Jobs []*JobSummary
ResultSetSize uint64
}
// ListJobs returns the list of jobs you currently have.
func (s *JobClient) List(ctx context.Context, input *ListJobsInput) (*ListJobsOutput, error) {
fullPath := path.Join("/", s.client.AccountName, "jobs")
query := &url.Values{}
if input.RunningOnly {
query.Set("state", "running")
}
if input.Limit != 0 {
query.Set("limit", strconv.FormatUint(input.Limit, 10))
}
if input.Marker != "" {
query.Set("manta_path", input.Marker)
}
reqInput := client.RequestInput{
Method: http.MethodGet,
Path: fullPath,
Query: query,
}
respBody, respHeader, err := s.client.ExecuteRequestStorage(ctx, reqInput)
if respBody != nil {
defer respBody.Close()
}
if err != nil {
return nil, errors.Wrap(err, "unable to list jobs")
}
var results []*JobSummary
for {
current := &JobSummary{}
decoder := json.NewDecoder(respBody)
if err = decoder.Decode(&current); err != nil {
if err == io.EOF {
break
}
return nil, errors.Wrap(err, "unable to decode list jobs response")
}
results = append(results, current)
}
output := &ListJobsOutput{
Jobs: results,
}
resultSetSize, err := strconv.ParseUint(respHeader.Get("Result-Set-Size"), 10, 64)
if err == nil {
output.ResultSetSize = resultSetSize
}
return output, nil
}
// GetJobInput represents parameters to a GetJob operation.
type GetJobInput struct {
JobID string
}
// GetJobOutput contains the outputs of a GetJob operation.
type GetJobOutput struct {
Job *Job
}
// GetJob returns the list of jobs you currently have.
func (s *JobClient) Get(ctx context.Context, input *GetJobInput) (*GetJobOutput, error) {
fullPath := path.Join("/", s.client.AccountName, "jobs", input.JobID, "live", "status")
reqInput := client.RequestInput{
Method: http.MethodGet,
Path: fullPath,
}
respBody, _, err := s.client.ExecuteRequestStorage(ctx, reqInput)
if respBody != nil {
defer respBody.Close()
}
if err != nil {
return nil, errors.Wrap(err, "unable to get job")
}
job := &Job{}
decoder := json.NewDecoder(respBody)
if err = decoder.Decode(&job); err != nil {
return nil, errors.Wrap(err, "unable to decode get job response")
}
return &GetJobOutput{
Job: job,
}, nil
}
// GetJobOutputInput represents parameters to a GetJobOutput operation.
type GetJobOutputInput struct {
JobID string
}
// GetJobOutputOutput contains the outputs for a GetJobOutput operation. It is your
// responsibility to ensure that the io.ReadCloser Items is closed.
type GetJobOutputOutput struct {
ResultSetSize uint64
Items io.ReadCloser
}
// GetJobOutput returns the current "live" set of outputs from a job. Think of
// this like `tail -f`. If error is nil (i.e. the operation is successful), it is
// your responsibility to close the io.ReadCloser named Items in the output.
func (s *JobClient) GetOutput(ctx context.Context, input *GetJobOutputInput) (*GetJobOutputOutput, error) {
fullPath := path.Join("/", s.client.AccountName, "jobs", input.JobID, "live", "out")
reqInput := client.RequestInput{
Method: http.MethodGet,
Path: fullPath,
}
respBody, respHeader, err := s.client.ExecuteRequestStorage(ctx, reqInput)
if respBody != nil {
defer respBody.Close()
}
if err != nil {
return nil, errors.Wrap(err, "unable to get job output")
}
output := &GetJobOutputOutput{
Items: respBody,
}
resultSetSize, err := strconv.ParseUint(respHeader.Get("Result-Set-Size"), 10, 64)
if err == nil {
output.ResultSetSize = resultSetSize
}
return output, nil
}
// GetJobInputInput represents parameters to a GetJobOutput operation.
type GetJobInputInput struct {
JobID string
}
// GetJobInputOutput contains the outputs for a GetJobOutput operation. It is your
// responsibility to ensure that the io.ReadCloser Items is closed.
type GetJobInputOutput struct {
ResultSetSize uint64
Items io.ReadCloser
}
// GetJobInput returns the current "live" set of inputs from a job. Think of
// this like `tail -f`. If error is nil (i.e. the operation is successful), it is
// your responsibility to close the io.ReadCloser named Items in the output.
func (s *JobClient) GetInput(ctx context.Context, input *GetJobInputInput) (*GetJobInputOutput, error) {
fullPath := path.Join("/", s.client.AccountName, "jobs", input.JobID, "live", "in")
reqInput := client.RequestInput{
Method: http.MethodGet,
Path: fullPath,
}
respBody, respHeader, err := s.client.ExecuteRequestStorage(ctx, reqInput)
if respBody != nil {
defer respBody.Close()
}
if err != nil {
return nil, errors.Wrap(err, "unable to get job input")
}
output := &GetJobInputOutput{
Items: respBody,
}
resultSetSize, err := strconv.ParseUint(respHeader.Get("Result-Set-Size"), 10, 64)
if err == nil {
output.ResultSetSize = resultSetSize
}
return output, nil
}
// GetJobFailuresInput represents parameters to a GetJobFailures operation.
type GetJobFailuresInput struct {
JobID string
}
// GetJobFailuresOutput contains the outputs for a GetJobFailures operation. It is your
// responsibility to ensure that the io.ReadCloser Items is closed.
type GetJobFailuresOutput struct {
ResultSetSize uint64
Items io.ReadCloser
}
// GetJobFailures returns the current "live" set of outputs from a job. Think of
// this like `tail -f`. If error is nil (i.e. the operation is successful), it is
// your responsibility to close the io.ReadCloser named Items in the output.
func (s *JobClient) GetFailures(ctx context.Context, input *GetJobFailuresInput) (*GetJobFailuresOutput, error) {
fullPath := path.Join("/", s.client.AccountName, "jobs", input.JobID, "live", "fail")
reqInput := client.RequestInput{
Method: http.MethodGet,
Path: fullPath,
}
respBody, respHeader, err := s.client.ExecuteRequestStorage(ctx, reqInput)
if respBody != nil {
defer respBody.Close()
}
if err != nil {
return nil, errors.Wrap(err, "unable to get job failures")
}
output := &GetJobFailuresOutput{
Items: respBody,
}
resultSetSize, err := strconv.ParseUint(respHeader.Get("Result-Set-Size"), 10, 64)
if err == nil {
output.ResultSetSize = resultSetSize
}
return output, nil
}