Hello Stratos, >> After fighting with a couple of ceph/rbd bugs, today I managed to set >> up a fully operational ceph cluster. Instance creation works, but >> burnin does not: can you please tell me which commit ID did you base >> your tests on? I want to make sure I debug an RBD issue and not >> something unrelated. > > Concerning rbd/rados, we encountered a couple of issues too, while > testing RBD disks with Ganeti, and so we cherry-picked some commits from > the master branch, and applied them on v0.39. However, since v0.40 will > be probably released soon, and will contain a lot of reworked/refactored > code, I think it would make sense to build and test Ceph from the master > branch, instead of debugging v0.39 code. > > The most critical issues we encountered and their fixes/commits are > listed below:
Yes, I have already had #4 (that I "fixed" for testing by truncating to 20 chars the image name in ganeti); I am running ceph 0.39 and will switch to git master if I encounter some of these problems: thanks for the suggestion. > I should also note, that we use ext4 (and not btrfs) as the underlying > filesystem for the OSDs. Why are you using ext4? For what I understand, it seems like rados is designed to take advantage of btrfs features. Also, can you please tell me the Ganeti revision on which you based your patch? >> I will post a detailed patch review after testing it, but before >> please fix the issues that "make lint" will give you (there are a >> couple of them). > > Constantinos will follow-up with an e-mail about the RBD disk template > patchset. Great! Thanks, Andrea