luxolus/open-nomad
2
0
Fork 0
Code Issues 1 Pull requests Packages Releases Wiki Activity

tr: Fetch Wait channel before killTask in restart

Browse source 
Currently, if killTask results in the termination of a process before
calling WaitTask, Restart() will incorrectly return a TaskNotFound
error when using the raw_exec driver on Windows.
This commit is contained in:
Danielle Lancashire 2019-06-26 15:20:57 +02:00
parent 448b759578
commit b9ac184e1f
No known key found for this signature in database
GPG key ID: 8D65584EF3DDF91B
1 changed files with 7 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

13
client/allocrunner/taskrunner/lifecycle.go
View file

@ -28,18 +28,19 @@ func (tr *TaskRunner) Restart(ctx context.Context, event *structs.TaskEvent, fai
// Tell the restart tracker that a restart triggered the exit // Tell the restart tracker that a restart triggered the exit
tr.restartTracker.SetRestartTriggered(failure) tr.restartTracker.SetRestartTriggered(failure)
// Grab a handle to the wait channel that will timeout with context cancelation
// _before_ killing the task.
waitCh, err := handle.WaitCh(ctx)
if err != nil {
return err
}
// Kill the task using an exponential backoff in-case of failures. // Kill the task using an exponential backoff in-case of failures.
if err := tr.killTask(handle); err != nil { if err := tr.killTask(handle); err != nil {
// We couldn't successfully destroy the resource created. // We couldn't successfully destroy the resource created.
tr.logger.Error("failed to kill task. Resources may have been leaked", "error", err) tr.logger.Error("failed to kill task. Resources may have been leaked", "error", err)
} }
// Drain the wait channel or wait for the request context to be canceled
waitCh, err := handle.WaitCh(ctx)
if err != nil {
return err
}
select { select {
case <-waitCh: case <-waitCh:
case <-ctx.Done(): case <-ctx.Done():