scripts
scripts copied to clipboard
jenkins/kola/do: use gzipped image file
jenkins/kola/do: use gzipped image file
The bzipped image file sometimes results in a timeout during decompression on the remote side. Gzipped images are supposed to be better supported by DigitalOcean, so use those.
How to use
[ describe what reviewers need to do in order to validate this PR ]
Testing done
[Describe the testing you have done before submitting this PR. Please include both the commands you issued as well as the output you got.]
- [ ] Changelog entries added in the respective
changelog/directory (user-facing change, bug fix, security fix, update)
TODO: update https://github.com/flatcar-linux/flatcar-docs/blob/main/docs/installing/cloud/digitalocean.md as soon as the gzipped DigitalOcean images make it to all the channels.
Maybe we should backport the change now to align the build system in all channels. So far we had no problems coming up from the change, and misalignment always causes inconsistencies (e.g., the gzip change gets documented but LTS doesn't have it) and cherry-pick troubles later.
The changes to compression handling? Yeah, we can do that. The reason this PR is not merged is because when I started testing it, the import still failed with a timeout several times in a row. So I lost some faith in whether bzip/gzip makes a real difference. Had to bump up the timeout instead (https://github.com/flatcar-linux/mantle/commit/d48d25de1cd3ba932ca63e7ef1f03a260f524762).
I think it was even a problem without any timeout constraints but maybe it doesn't happen often anymore because the hardware running the decompression job got upgraded as well or has less load than before, or lbzip2 is used there now…
I'm still for merging but think that if we do so we should first cherry-pick everywhere.
We can merge this now and cherry pick
Merging this as is now is pointless as this updates scripts in jenkins directory, which are rather unused.
Ah, we now need to change DIGITALOCEAN_IMAGE_URL_TEMPLATE in ci-automation/ci-config.env
I'm still not convinced this helps. When testing this I found that there were still occasional timeouts during the upload, same as when using bzip2. If no one objects I'd rather close.
The main benefit would be a faster image import when running the tests
If it were faster I would agree, but I don't remember seeing any difference. I still encourage someone else to give it a try.