search

TheJacksonLaboratory / splicing-pipelines-nf

Repository for the Anczukow-Lab splicing pipeline
14 stars 9 forks source link

Using new STAR index fails on CloudOS GCP env #282

Closed Vlad-Dembrovskyi closed 2 years ago

Vlad-Dembrovskyi commented 2 years ago

Problem

Using new STAR index (v2.7.9a) with latest dev is working fine locally, but fails on CloudOS: https://cloudos.lifebit.ai/public/jobs/615734ae87389401cf6b9f07

Solution

1) Check if the uploaded index is correct; 2) Check if the correct container is used.

Vlad-Dembrovskyi commented 2 years ago

The reason was that in google.config there were old docker images specified for many processes, including STAR process. And since profiles in our case are called after the process scope in nextflow config, the containers in google.config overwrote those in nextflow.config. Old containers cannot take the new index, and that's why the pipeline actually failed. It was not possible to reproduce locally, because locally you don't use the google.config (unless you explicitly include it and force set executor to local, or false). And it was impossible to debug on cloudos, because the workdirs on GCP cloudos cannot be opened to view the logs (known issue https://lifebit.atlassian.net/browse/CQA-425)