fix compose up wait for state #553
Conversation
| for _, serviceInfo := range serviceInfos { | ||
| serviceList = append(serviceList, compose.NormalizeServiceName(serviceInfo.Service.Name)) | ||
| } | ||
| var errFailedToReachStartedState = errors.New("failed to reach STARTED state") |
There was a problem hiding this comment.
declare errors where they are returned
| } | ||
|
|
||
| serviceState[newStatus.Name] = newStatus.State | ||
|
|
||
| if allInState(targetState, serviceState) { | ||
| for _, sInfo := range serviceInfos { |
There was a problem hiding this comment.
this was unused, so I removed it
There was a problem hiding this comment.
Added back to commands.go in #562, as the value is used in printEndpoints
| @@ -38,14 +37,6 @@ func Subscribe(ctx context.Context, client client.Client, services []string) (<- | |||
| defer serverStream.Close() | |||
| defer close(statusChan) | |||
| for { | |||
| // handle cancel from caller; TODO: do we need this? Receive() should return false on context done | |||
There was a problem hiding this comment.
see comment; wasn't needed
| // handle cancel from caller; TODO: do we need this? Receive() should return false on context done | ||
| select { | ||
| case <-ctx.Done(): | ||
| term.Debug("Context Done - exiting Subscribe goroutine") |
There was a problem hiding this comment.
avoid printing in go-routine since it will mess up other prints in main program
There was a problem hiding this comment.
This is more of a failure of our logging package than requirement, but I agree we should keep where logging happens, and try to avoid spread logging into all layers of the dependencies.
| @@ -56,7 +47,6 @@ func Subscribe(ctx context.Context, client client.Client, services []string) (<- | |||
| } | |||
| continue | |||
| } | |||
| term.Debug("Subscribe Stream closed", serverStream.Err()) | |||
There was a problem hiding this comment.
ditto; prints moved to caller
| if subStatus.Name == "" && (servInfo != nil && servInfo.Service != nil) { | ||
| subStatus.Name = servInfo.Service.Name | ||
| subStatus.Status = servInfo.Status | ||
| subStatus.State = convertServiceState(servInfo.Status) | ||
| } | ||
|
|
||
| statusChan <- subStatus | ||
| term.Debugf("service %s with state ( %s ) and status: %s\n", subStatus.Name, subStatus.State.String(), subStatus.Status) |
lionello
force-pushed
the
lio-fix-compose-up
branch
from
683e927 to
479a908
Compare
|
|
||
| // blocking call to tail | ||
| if err := cli.Tail(ctx, client, tailParams); err != nil { | ||
| if errors.Is(context.Cause(ctx), errDeploymentFailed) { | ||
| term.Warn("Deployment FAILED. Service(s) not running.") | ||
|
|
||
| if !nonInteractive { |
There was a problem hiding this comment.
non-interactive prompt in main thread now, which fixes the issue @raphaeltm was having
Refactor the async code. Please compare with #518, I think this one's cleaner.
We should introduce a build failure state to also trigger SERVICE_FAILED so we'll get AI-debugger for build failures. This requires some code in Fabric (probably needs to handle kaniko STOPPED state)