timeout: 0 will fail the pipeline immediately

[lafh]

Expected Behavior

timeout: "0" sets no timeout for PipelineRun or TaskRun.

Actual Behavior

apiVersion: tekton.dev/v1alpha1
kind: PipelineRun
metadata:
  name: pipelin-run
spec:
  ...
  timeout: "0"

failed immediately on the HEAD release.

[imjasonh]

I believe this is working as intended, though we should probably validate that a timeout is >0 (or > some reasonable minimum) if it's specified.

If you don't specify a timeout, the controller will set a default timeout, which is currently one hour.

There is not currently a way to specify "this *Run should be allowed to run indefinitely forever" since, the thinking goes, that's usually not what you actually want. Generally you have some idea that this process should finish in some bounded amount of time.

Bounded timeouts can be useful to operators, since it allows them to block new incoming runs and calculate how long at most they'll have to wait for all existing runs to finish. This can be especially useful when migrating or upgrading workloads, without having to forcibly cancel anything.

[lafh]

It seems "no timeout" is an option according to #978 and the documentation https://github.com/tektoncd/pipeline/blob/master/docs/pipelineruns.md

"this *Run should be allowed to run indefinitely forever" is unlikely to happen, but "no timeout" should still be an option for flexibility reason. For example, we might want to control the timeout at a higher level (e.g. by killing the pipelines instead of it failing automatically).

[imjasonh]

Ah okay, I must have been mistaken. If the docs says timeout: 0 should work and it doesn't that's a bug. Reopening!

[imjasonh]

@houshengbo Can you look into this and see if you can reproduce? (I ask because you implemented #1040)