uniqueness of the task podname [k8s-platform]

[naineshmp]

I am exploring the config k8s.cleanup which skips podCleanup after successful execution of the task.

It seems pod.name is set as task.hash and task.hash seems to depend on inputs, session.id, def, name etc.

Question - If I have two pipeline steps, with similar definition and inputs, and is there a clash for pod-name? If I run these two tasks in serial, only one of them will be successful? Sorry, if this is already explained somewhere. I am just curious about uniqueness of task-pod-name.

Currently, pod_name (hash) seems to be using multiple keys [ session.uniqueId, name, task.source ] + inputs

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

Lines 2015 to 2022 in 41021cb

	final protected HashCode createTaskHashKey(TaskRun task) {
	List keys = [ session.uniqueId, name, task.source ]
	if( task.isContainerEnabled() )
	keys << task.getContainerFingerprint()
	// add all the input name-value pairs to the key generator

[pditommaso]

Each task is associated with a hash identity computed using the task inputs, the command to be executed and the pipeline execution unique identifier (session-id). Therefore two identical tasks in different pipelines will get a different hash id -> pod name