osbuild: switch compression to off; workaround image corruption #3730
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Raw format is fine to use on systems that have reflinks for example.
I've been investigating why a seemingly innocuous change
(changing compression on OSBuild generated qemu qcow2) would
cause disk images to not boot [1]. I think I have found the issue.
I was first trying to make sure 100% that the files got written
out over the virtiofs mount before the VM got shutdown so I decided
to add a `umount $workdir` to the process. But this ended up with
a `umount: /srv/: target is busy.` error.
When the supermin VM gets run we `cd "${workdir}"` at the end of
supermin-init-prelude.sh. This has the effect of causing all
spawned processes (including PID1/init) to have a cwd of /srv/.
```
bash-5.2# lsof /srv
COMMAND PID USER FD TYPE DEVICE SIZE/OFF NODE NAME
init 1 root cwd DIR 0,26 4096 10485829 /srv
kthreadd 2 root cwd DIR 0,26 4096 10485829 /srv
pool_work 3 root cwd DIR 0,26 4096 10485829 /srv
kworker/R 4 root cwd DIR 0,26 4096 10485829 /srv
...
...
```
Which means it's unlikely that the virtiofs mount ever gets cleanly
unmounted. Let's rework things here so that actual work gets spawned
in a subshell to prevent `init` from having a cwd on the virtiofs mount.
We also add in an `umount` of the cache qcow2 (if exists) and the virtiofs
mount to strengthen our chances of a clean unmount.
[1] coreos#3728
Just in case this is what is causing issues with file consistency when copying out of the supermin VM.
And drop all patches that have now been upstreamed. The only remaining patches are one to enable s390x builds to work while we figure out [1] and another that adds a log statement when cache eviction happens, which I plan to upstream at some point. [1] coreos/fedora-coreos-tracker#1667
This will compare the image that was just created to see if it has any problems.
We think there might be some XFS reflink issues when we run the OSBuild org.osbuild.qemu stage compression: false. See coreos#3728 (comment)
We previously did this in a different way (2a8d1e6) but then had to revert it (39fdd61) because it caused images to not boot [1]. The root cause appears to have been the virtiofs mount not being unmounted cleanly from the supermin VM and that is now fixed so let's switch back to not compressing since we rely on our outer compression [2]. [1] coreos#3728 [2] coreos/fedora-coreos-tracker#1653 (comment)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This obsoletes #3729 but keeps a lot of the cleanups done there to make unmounting cleaner. This biggest change here is switching the cache.qcow2 to ext4 (see #3728 (comment)) and adding a sanity check in OSBuild to catch if the created image is ever inconsistent.
This also now updates OSBuild to v108 and drops many patches as a result.