Details
-
Type:
Bug
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: jenkins
-
Labels:None
-
Team:SQuaRE
-
Urgent?:No
Description
This package is being successfully built, rebuilt from source as part of the pipelines/release/tarball.groovy job, and even smoke-tested from the temporary filesystem "distribution server". I see the appropriate tarball in the distrib directory in /project/jenkins/prod at NCSA. The aws s3 cp command from that distrib directory appears to succeed, but the tarball does not end up being visible in S3.
I'm wondering if this is due to the S3 bucket filling up.
Attachments
Activity
/project/jenkins/prod/agent-ldfc-ws-2/ws/release/tarball/b09919c18f/distrib at NCSA is the directory that is being copied from. It contains obs_base-g20ff5da1ef+ff9f4b6609@Linux64.tar.gz, which is the tarball that is to be copied. That should end up at https://eups.lsst.codes/stack/redhat/el7/conda-system/miniconda3-py38_4.9.2-0.7.0/ but it is not there.
[slack summary: we have established eups.lsst.codes serves from a PV on Google...]
I updated the disk size to 4096 GiB and the PVC to request 4 TiB, and things seem to be working again. The synchronization process from AWS S3 copied over the missing tarball even without rerunning a publish job in Jenkins.
https://ci.lsst.codes/blue/organizations/jenkins/release%2Ftarball/detail/tarball/7762/pipeline is the publish Jenkins job with the supposedly-successful copy.