chanzuckerberg / czid-workflows

Portable WDL workflows for CZ ID production pipelines
https://czid.org/
MIT License
37 stars 7 forks source link

Re-Attach to running alignment chunks on retry #372

Closed morsecodist closed 3 months ago

morsecodist commented 4 months ago

I went through a few iterations of this but I think the simplest way is to base the approach on a cache. We don't want to make the same batch call twice. The first batch call puts a job on the spot queue, the second the on-demand queue. Also if for whatever reason we change the job definition or inputs that is also part of the cache key.

I also added the convention of adding the tag AlignmentCoordination=True to these files so we can clean them with a lifecycle policy. More for cleanliness than cost. I am not sure what convention you are using. Does this make sense @rzlim08 ?