Open boegel opened 1 year ago
Left container; tar'ing up /tmp/bot/EESSI/eessi.BM69QGLjmn for future inspection ... created tarball '/mnt/shared/home/bot/eessi-bot-software-layer/jobs/2023.06/pr_188/event_bf958ee0-0a82-11ee-93ec-8e0bfaa03c26/run_001/linux_aarch64_generic/eessi-2023.06-compat/job_4940_1686744594.tgz' To resume work add '--resume /tmp/bot/EESSI/eessi.BM69QGLjmn' Directory //cvmfs/pilot.eessi-hpc.org/versions/2023.06 was not created, not creating tarball.
The `job_*.tgz` mentioned not only includes `/cvmfs/...`, but also `/tmp` and `apptainer_cache`, so it can't be used for ingestion.
As the first line states, this tarball is intended for future inspection not for ingestion. It's essentially the complete temporary directory used during the job or more precisely what install_compatibility_layer.sh
uses as base folder for temporary data. It includes cvmfs
, home
, tmp
and possibly container_cachedir
.
What's missing in the bot/build.sh
script is assigning a value to eessi_tmpdir
. Adding a line such as
eessi_tmpdir=${STORAGE}
before https://github.com/EESSI/compatibility-layer/blob/114f12ab6806232ea7b894ac5af73aeb7d2443fa/bot/build.sh#L97-L106 would probably fix it.
As shown in #188, building the
2023.06
compat layer autonomously worked, but creating the tarball for ingestion failed because of a bug at the end of thebot/build.sh
script.Relevant part of output of the Slurm build job:
The
job_*.tgz
mentioned not only includes/cvmfs/...
, but also/tmp
andapptainer_cache
, so it can't be used for ingestion.