Locked by another command is not visible as pending

[n0rad]

Hello,

While using github, when a plan is in progress and another plan is triggered, manually or especially by a new commit when iterating on the PR, Atlantis send the following comment :

The default workspace is currently locked by another command that is running for this pull request.
Wait until the previous command is complete and try again.

and set the check status as failed.

it's disturbing because the user cannot be sure that something is really still in progress, depending on when he is browsing the PR.
I think Atlantis should set the check as pending, so it's clearly visible that the user should wait.

[jamengual]

is this still happening with v0.19.8?

[n0rad]

just tested on 0.19.7 and it's still the case. I will check on 0.19.8 when possible.

[jhaumont]

I face the same issue today ! Any update?

[github-actions]

This issue is stale because it has been open for 1 month with no activity. Remove stale label or comment or this will be closed in 1 month.'

[DavidPessoa]

I have this issue on 0.22.3
The default workspace at path .... is currently locked by another command that is running for this pull request.
Wait until the previous command is complete and try again.

With parallel plan/apply disabled

[github-actions]

This issue is stale because it has been open for 1 month with no activity. Remove stale label or comment or this will be closed in 1 month.'

[nitrocode]

The message comes from TryLock function

atlantis/server/events/working_dir_locker.go

Lines 77 to 95 in dfe6ee0

	func (d *DefaultWorkingDirLocker) TryLock(repoFullName string, pullNum int, workspace string, path string) (func(), error) {
	d.mutex.Lock()
	defer d.mutex.Unlock()
	pullKey := d.pullKey(repoFullName, pullNum)
	workspaceKey := d.workspaceKey(repoFullName, pullNum, workspace, path)
	for _, l := range d.locks {
	if l == pullKey || l == workspaceKey {
	return func() {}, fmt.Errorf("The %s workspace at path %s is currently locked by another"+
	" command that is running for this pull request.\n"+
	"Wait until the previous command is complete and try again.", workspace, path)
	}
	}
	d.locks = append(d.locks, workspaceKey)
	return func() {
	d.unlock(repoFullName, pullNum, workspace, path)
	}, nil
	}

Here are all the references to that function

https://github.com/search?q=repo%3Arunatlantis%2Fatlantis+TryLock&type=code

I'd imagine it all starts at the pre workflow run's RunPreHooks function

atlantis/server/events/pre_workflow_hooks_command_runner.go

Lines 60 to 65 in dfe6ee0

	unlockFn, err := w.WorkingDirLocker.TryLock(baseRepo.FullName, pull.Num, DefaultWorkspace, DefaultRepoRelDir)
	if err != nil {
	return err
	}
	log.Debug("got workspace lock")
	defer unlockFn()

Which is then called from these locations... But from looking at these, it seems it catches the error, skips over them, and runs the command anyway

atlantis/server/events/command_runner.go

Lines 166 to 174 in dfe6ee0

	err = c.PreWorkflowHooksCommandRunner.RunPreHooks(ctx, nil)
	if err != nil {
	ctx.Log.Err("Error running pre-workflow hooks %s. Proceeding with %s command.", err, command.Plan)
	}
	autoPlanRunner := buildCommentCommandRunner(c, command.Plan)
	autoPlanRunner.Run(ctx, nil)

atlantis/server/events/command_runner.go

Lines 290 to 298 in dfe6ee0

	err = c.PreWorkflowHooksCommandRunner.RunPreHooks(ctx, cmd)
	if err != nil {
	ctx.Log.Err("Error running pre-workflow hooks %s. Proceeding with %s command.", err, cmd.Name.String())
	}
	cmdRunner := buildCommentCommandRunner(c, cmd.CommandName())
	cmdRunner.Run(ctx, cmd)

More code diving is needed. This is to get someone started.

[github-actions]

This issue is stale because it has been open for 1 month with no activity. Remove stale label or comment or this will be closed in 1 month.'