Skip to content
Permalink
master
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Go to file
 
 
Cannot retrieve contributors at this time
// Copyright 2019 Drone IO, Inc.
//
// 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 manager
import (
"context"
"encoding/json"
"time"
"github.com/drone/drone/core"
"github.com/drone/drone/store/shared/db"
"github.com/drone/go-scm/scm"
"github.com/hashicorp/go-multierror"
"github.com/sirupsen/logrus"
)
type teardown struct {
Builds core.BuildStore
Events core.Pubsub
Logs core.LogStream
Scheduler core.Scheduler
Repos core.RepositoryStore
Steps core.StepStore
Status core.StatusService
Stages core.StageStore
Users core.UserStore
Webhook core.WebhookSender
}
func (t *teardown) do(ctx context.Context, stage *core.Stage) error {
logger := logrus.WithField("stage.id", stage.ID)
logger.Debugln("manager: stage is complete. teardown")
build, err := t.Builds.Find(noContext, stage.BuildID)
if err != nil {
logger.WithError(err).Warnln("manager: cannot find the build")
return err
}
logger = logger.WithFields(
logrus.Fields{
"build.number": build.Number,
"build.id": build.ID,
"repo.id": build.RepoID,
},
)
repo, err := t.Repos.Find(noContext, build.RepoID)
if err != nil {
logger.WithError(err).Warnln("manager: cannot find the repository")
return err
}
for _, step := range stage.Steps {
if len(step.Error) > 500 {
step.Error = step.Error[:500]
}
err := t.Steps.Update(noContext, step)
if err != nil {
logger.WithError(err).
WithField("stage.status", stage.Status).
WithField("step.name", step.Name).
WithField("step.id", step.ID).
Warnln("manager: cannot persist the step")
return err
}
}
if len(stage.Error) > 500 {
stage.Error = stage.Error[:500]
}
stage.Updated = time.Now().Unix()
err = t.Stages.Update(noContext, stage)
if err != nil {
logger.WithError(err).
Warnln("manager: cannot update the stage")
return err
}
for _, step := range stage.Steps {
t.Logs.Delete(noContext, step.ID)
}
stages, err := t.Stages.ListSteps(noContext, build.ID)
if err != nil {
logger.WithError(err).Warnln("manager: cannot get stages")
return err
}
//
//
//
err = t.cancelDownstream(ctx, stages)
if err != nil {
logger.WithError(err).
Errorln("manager: cannot cancel downstream builds")
return err
}
err = t.scheduleDownstream(ctx, stage, stages)
if err != nil {
logger.WithError(err).
Errorln("manager: cannot schedule downstream builds")
return err
}
//
//
//
if isBuildComplete(stages) == false {
logger.Debugln("manager: build pending completion of additional stages")
return nil
}
logger.Debugln("manager: build is finished, teardown")
build.Status = core.StatusPassing
build.Finished = time.Now().Unix()
for _, sibling := range stages {
if sibling.Status == core.StatusKilled {
build.Status = core.StatusKilled
break
}
if sibling.Status == core.StatusFailing {
build.Status = core.StatusFailing
break
}
if sibling.Status == core.StatusError {
build.Status = core.StatusError
break
}
}
if build.Started == 0 {
build.Started = build.Finished
}
err = t.Builds.Update(noContext, build)
if err == db.ErrOptimisticLock {
logger.WithError(err).
Warnln("manager: build updated by another goroutine")
return nil
}
if err != nil {
logger.WithError(err).
Warnln("manager: cannot update the build")
return err
}
repo.Build = build
repo.Build.Stages = stages
data, _ := json.Marshal(repo)
err = t.Events.Publish(noContext, &core.Message{
Repository: repo.Slug,
Visibility: repo.Visibility,
Data: data,
})
if err != nil {
logger.WithError(err).
Warnln("manager: cannot publish build event")
}
payload := &core.WebhookData{
Event: core.WebhookEventBuild,
Action: core.WebhookActionUpdated,
Repo: repo,
Build: build,
}
err = t.Webhook.Send(noContext, payload)
if err != nil {
logger.WithError(err).Warnln("manager: cannot send global webhook")
}
user, err := t.Users.Find(noContext, repo.UserID)
if err != nil {
logger.WithError(err).
Warnln("manager: cannot find repository owner")
// this error is insufficient to fail the function,
// however, execution of the function should be halted
// to prevent a nil pointer in subsequent operations.
return nil
}
req := &core.StatusInput{
Repo: repo,
Build: build,
}
err = t.Status.Send(noContext, user, req)
if err != nil && err != scm.ErrNotSupported {
logger.WithError(err).
Warnln("manager: cannot publish status")
}
return nil
}
// cancelDownstream is a helper function that tests for
// downstream stages and cancels them based on the overall
// pipeline state.
func (t *teardown) cancelDownstream(
ctx context.Context,
stages []*core.Stage,
) error {
failed := false
for _, s := range stages {
if s.IsFailed() {
failed = true
}
}
var errs error
for _, s := range stages {
if s.Status != core.StatusWaiting {
continue
}
var skip bool
if failed == true && s.OnFailure == false {
skip = true
}
if failed == false && s.OnSuccess == false {
skip = true
}
if skip == false {
continue
}
if areDepsComplete(s, stages) == false {
continue
}
logger := logrus.WithFields(
logrus.Fields{
"stage.id": s.ID,
"stage.on_success": s.OnSuccess,
"stage.on_failure": s.OnFailure,
"stage.is_failure": failed,
"stage.depends_on": s.DependsOn,
},
)
logger.Debugln("manager: skipping step")
s.Status = core.StatusSkipped
s.Started = time.Now().Unix()
s.Stopped = time.Now().Unix()
err := t.Stages.Update(noContext, s)
if err == db.ErrOptimisticLock {
t.resync(ctx, s)
continue
}
if err != nil {
logger.WithError(err).
Warnln("manager: cannot update stage status")
errs = multierror.Append(errs, err)
}
}
return errs
}
// scheduleDownstream is a helper function that tests for
// downstream stages and schedules stages if all dependencies
// and execution requirements are met.
func (t *teardown) scheduleDownstream(
ctx context.Context,
stage *core.Stage,
stages []*core.Stage,
) error {
var errs error
for _, sibling := range stages {
if sibling.Status == core.StatusWaiting {
if len(sibling.DependsOn) == 0 {
continue
}
// PROBLEM: isDep only checks the direct parent
// i think ....
// if isDep(stage, sibling) == false {
// continue
// }
if areDepsComplete(sibling, stages) == false {
continue
}
// if isLastDep(stage, sibling, stages) == false {
// continue
// }
logger := logrus.WithFields(
logrus.Fields{
"stage.id": sibling.ID,
"stage.name": sibling.Name,
"stage.depends_on": sibling.DependsOn,
},
)
logger.Debugln("manager: schedule next stage")
sibling.Status = core.StatusPending
sibling.Updated = time.Now().Unix()
err := t.Stages.Update(noContext, sibling)
if err == db.ErrOptimisticLock {
t.resync(ctx, sibling)
continue
}
if err != nil {
logger.WithError(err).
Warnln("manager: cannot update stage status")
errs = multierror.Append(errs, err)
}
err = t.Scheduler.Schedule(noContext, sibling)
if err != nil {
logger.WithError(err).
Warnln("manager: cannot schedule stage")
errs = multierror.Append(errs, err)
}
}
}
return errs
}
// resync updates the stage from the database. Note that it does
// not update the Version field. This is by design. It prevents
// the current go routine from updating a stage that has been
// updated by another go routine.
func (t *teardown) resync(ctx context.Context, stage *core.Stage) error {
updated, err := t.Stages.Find(ctx, stage.ID)
if err != nil {
return err
}
stage.Status = updated.Status
stage.Error = updated.Error
stage.ExitCode = updated.ExitCode
stage.Machine = updated.Machine
stage.Started = updated.Started
stage.Stopped = updated.Stopped
stage.Created = updated.Created
stage.Updated = updated.Updated
return nil
}