Strange behaviour of cache and resume with docker failures

[awgymer]

Bug report

I am not actually clear on whether this is a bug or whether this is some expected behaviour manifesting in a strange way to the end user.

When trying to help debug issues with seemingly valid caches for jobs which had failed the first attempt but succeeded on a retry I discovered that you can cause processes to not be retrieved from a cache in a weird manner.

It's somewhat hard to explain clearly so I attach a link to a repo which has a minimal example that can reproduce the observed behaviour.

https://github.com/awgymer/nextflow-cache-repository

[zachary-foster]

Here are logs from two runs of the pathogensurveillance pipeline where jobs that are retried and then succeed are not cached correctly. Not sure if this will be helpful or not. Keep in mind that these pipeline runs are ended prematurely at different points. One example of a process that does not cache correctly is the SPADES job with the cache eb2a022adaba6e264277b2974fbc5133 for sample "SRR16542347".

hash_log.txt
hash_log_2.txt