Azure Batch executor does not always respect error strategy

[raylim]

Bug report

Expected behavior and actual behavior

Error strategy should be respected, but when a process fails due to running out of walltime, the whole pipeline is terminated.

Steps to reproduce the problem

process TEST {
    executor 'azurebatch'
    container 'ubuntu'
    errorStrategy 'ignore'
    time "10s"
    script:
    """
    sleep 100
    """
}

workflow {
    TEST()
}

Program output

$ nextflow run main.nf  -w az://nextflow/test                                                                        [90/4787]

 N E X T F L O W   ~  version 24.08.0-edge

Launching `main.nf` [grave_hopper] DSL2 - revision: 8398d93773

executor >  azurebatch (1)
[6b/6fa6f5] TEST [100%] 1 of 1, failed: 1 ✘
ERROR ~ Error executing process > 'TEST'

Caused by:
  The task was ended by user request


Command executed:

  sleep 100

Command exit status:
  -

Command output:
  (empty)

Work dir:
  az://nextflow/test/6b/6fa6f56c51b72a15dde5893e4ac028

Tip: you can try to figure out what's wrong by changing to the process work dir and showing the script file named `.command.sh`

 -- Check '.nextflow.log' file for details

Environment

• Nextflow version: 24.08.0-edge build 5922
• Java version: openjdk 17.0.3-internal 2022-04-19
• Operating system: Linux
• Bash version: (GNU bash, version 4.4.19(1)-release (x86_64-redhat-linux-gnu)

[bentsherman]

How did the task fail? What does this mean?

Caused by:
  The task was ended by user request

[raylim]

I've also had this issue happen with the slurm executor. Error strategy works with the latest stable release so long as a closure isn't used.

[bentsherman]

It looks like certain errors cannot be retried or ignored:

nextflow/modules/nextflow/src/main/groovy/nextflow/processor/TaskProcessor.groovy

Lines 1064 to 1066 in 324de3d

	// -- when is a task level error and the user has chosen to ignore error,
	// just report and error message and DO NOT stop the execution
	if( task && error instanceof ProcessException ) {

nextflow/modules/nextflow/src/main/groovy/nextflow/processor/TaskProcessor.groovy

Lines 1099 to 1116 in 324de3d

	message << "Error executing process > '${safeTaskName(task)}'"
	switch( error ) {
	case ProcessException:
	formatTaskError( message, error, task )
	break
	case ProcessEvalException:
	formatCommandError( message, error, task )
	break
	case FailedGuardException:
	formatGuardError( message, error as FailedGuardException, task )
	break;
	default:
	message << formatErrorCause(error)
	dumpStackTrace = true
	}

Need to figure out which case is being reached here

[raylim]

An error executing a process will bring everything to a halt (on version 24.04.4):

Caused by:
  Status code 409, {
    "odata.metadata":"https://ocra.eastus2.batch.azure.com/$metadata#Microsoft.Azure.Batch.Protocol.Entities.Container.errors/@Element","code":"TaskExists","message":{
      "lang":"en-US","value":"The specified task already exists.\nRequestId:d1437934-d0cd-4924-9fb3-1b500406ec0b\nTime:2024-10-24T20:26:59.8426275Z"
    }
  }



 -- Check '.nextflow.log' file for details
WARN: Killing running tasks (11100)

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[bentsherman]

I wonder if this was resolved by #5806