fix: clear cache and retry on cvmfs installation failure #48
Conversation
| sudo rm -rf /var/cache/apt/archives | ||
| sudo rm -rf /var/cache/apt/lists | ||
| sudo apt-get -q update | ||
| sudo apt-get -q -y install cvmfs |
There was a problem hiding this comment.
I know this is unlikely to fail again, but you should probably check anyway given why you are introducing this change
There was a problem hiding this comment.
Yeah, I just don't think this PR is the right solution. If you look at your original failing pipeline: https://github.com/DUNE-DAQ/daq-release/actions/runs/16244930712/job/45866367226#step:3:58, you'll see that it is already running in bash -e, so doing something like if [ $? -ne 0 ] is not going to catch this since we probably don't even return a failing exit code from apt-get install (or we would have bailed out earlier).
I have been trying to find some way to trigger this issue (GHA or locally), but no luck so far.
|
What about caching ( by default ) only the debs, not the lists? I don't think we can hit this error in that case. Add the current behavior as an opt-in parameter, if you care about speed but don't mind to clear the cache once in a while. |
|
Maybe the solution here is not to use apt at all. We could either use our service container (once it supports autofs, should be soon) or build a statically linked cvmfs2 binary that bundles all dependencies. The latter would also make cvmfsexec significantly simpler. |
Can you really avoid all the dependencies with a static build? https://github.com/cvmfs-contrib/github-action-cvmfs/actions/runs/21126022183/job/60747366609#step:3:163 Container would be interesting to explore but likely a larger download than the current installation, which may make it slower over all. |
This PR fixes #43 by clearing the cache and retrying when
cvmfsfails to install.