New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Ceph bluestore #107822
base: master
Are you sure you want to change the base?
Ceph bluestore #107822
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
instead of the commit messages being ceph:
, i would do nixosTests.ceph:
as you're not affecting the underlying package, just the nixos tests.
Also, just copy the OSD bootstrap keyring to the OSD hosts instead of the admin keyring.
Is non-bluestore still a thing? If so, could we keep one non-bluestore node in the cluster? |
|
What I mean is, the test should probably test both, if we can easily. |
Yeah, seeing as at the OSD level, it really shouldn't matter what type of store is being used, testing them in the same test should definitively be doable. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
A thing we should probably think about is, if we want to add a ceph-volume test, which is basically just the single node test, but using ceph-volume in different ways. It all depends on how comprehensive we want to cover the ceph-volume tool honestly. Also a thing to be noted is that when using filestore, ceph-volume (at least before) required a journal partition to be defined.
The single-node test is still testing the "old style" non-LVM volume for osd0. I was originally planning to test ceph-volume with both filestore and bluestore, but I ran into the required separate journal too. Now that I think about it, that shouldn't be pretty easy to add to the single-node test. The multi-node test is currently only using bluestore because I wanted to test bluestore without the xfs module loaded, and I wasn't sure if I should refactor the node configurations to make them less uniform. How much do we need to worry about resource usage in these tests? I noticed they weren't blocking nixos-unstable updates from happening when ceph was broken, and I kind of want to write a more thorough / bigger HA cluster test, but I also wouldn't want to decrease the chances of tests getting included in regular hydra runs. |
ceph-volume insists on using an exernal XFS journal when using filestore.
I marked this as stale due to inactivity. → More info |
@aij any chance you can pick this up? As far as resource usage goes, even if tests can't be included into the regular hydra runs, having these tests, to validate version bumps to not break anything is still super valuable. |
Since bfc11c6 added bluestore tests, I'm not sure how useful this MR is any more. This does use Is it worth trying to update this, or should I just close it? I've been a bit low on time, but should have some time next week when I might be able to revisit it. |
Yeah, I wasn't aware bfc11c6 did land. In general, it seems the Maybe this could be rebased, updating It could even join the two files, introducing a |
I marked this as stale due to inactivity. → More info |
Motivation for this change
Test Ceph's Bluestore on NixOS and include necessary setup for
ceph-volume
to work.Things done
sandbox
innix.conf
on non-NixOS linux)nix-shell -p nixpkgs-review --run "nixpkgs-review wip"
./result/bin/
)nix path-info -S
before and after)