member badge
Adam Talbot
Currently, there is no way to set a maximum elapsed time for a pipeline, which would automatically cancel it once that time limit is reached. This can cause pipelines to enter an orphaned state where they keep running and using resources without being connected to Seqera Platform.
I would like to have a timeout option available on the pipeline launch page that does two things:
  • Sets the native API for timeouts on the executor (e.g., the job timeout on AWS Batch).
  • When that timeout is reached, Seqera Platform will cancel any related jobs to stop resource use in the executor.
This should have a default value depending on the compute environment, which can be adjusted per pipeline, e.g.,
24 hours
for the parent Compute Environment but optionally extended to
72 hours
for a pipeline that takes longer.
Using
-resume
should allow users to restart a pipeline that has timed out, so they do not lose too much time.