You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This workflow fails on Google Batch (with the process unexpectedly terminating, without any error), i.e., when using googlebatch profile, but succeeds locally, (with local profile).
I noticed, after looking at the nf-google code, and in particular
, that if instead I override the process containerOptions in the configuration, the workflow works on Google Batch. That is setting:
googlebatch {
process {
containerOptions = "-e VARIABLE_LOADED=true"
}
}
I believe this is kind of a serious bug. Also, it was extremely difficult to follow the entire process, as there is nowhere in .command.run anything about how the container is run, everything is being obfuscated in the following cryptic nxf_launch in .command.run, which just calls itself (?):
Bug report
Neither
nextflow.config
runOptions
andenvWhileList
arguments are considered when running on Google Batch.Expected behavior and actual behavior
To assess this, have a docker image with an entrypoint expecting an environment variable:
entrypoint.sh
and use this entrypoint in this example Dockerfile:
to build and push an image called
test
on${CONTAINERS_REPO}/${CONTAINERS_PROJECT_ID}/${CONTAINERS_REGISTRY}
.Then, have the following docker configuration in nextflow.config:
as well as the configuration required to run the process on Google Batch.
Finally, have this example process and workflow:
This workflow fails on Google Batch (with the process unexpectedly terminating, without any error), i.e., when using
googlebatch
profile, but succeeds locally, (withlocal
profile).I noticed, after looking at the nf-google code, and in particular
nextflow/plugins/nf-google/src/main/nextflow/cloud/google/batch/GoogleBatchTaskHandler.groovy
Line 243 in cc0163a
containerOptions
in the configuration, the workflow works on Google Batch. That is setting:I believe this is kind of a serious bug. Also, it was extremely difficult to follow the entire process, as there is nowhere in .command.run anything about how the container is run, everything is being obfuscated in the following cryptic nxf_launch in .command.run, which just calls itself (?):
Environment
The text was updated successfully, but these errors were encountered: