Ceph » CephFS

Download/install backuppc and get it into suite.

Ekapol Rojpiboonphun wrote:

Just to make sure that I will be on this along the line of what you might already have in mind. (More details please :)

Do you have the list of what we should back up from our ceph cluster? As part of our QA suite, I figure they should be the configuration, the scripts that we run, test results, test logs. Anything else that you might want to add to the list?

Actually, the point isn't to back up anything... it's to generate an interesting workload. There have been user reports in the past of MDS bugs triggered by backuppc, and we'd like to be able to trigger them. Most likely it is related to backuppc's heavy use of hard links.

What this really would be is a bash script (say, ceph.git/qa/workunits/suites/backuppc.sh) that will run backuppc several times and back up to the local directory. (Workunits are run from the CWD of a ceph mount, either the kernel client or ceph-fuse or whatever). I'd suggest backing up /usr and /var a few times. It may be that we want to also make the data backed up change more so there is more of an incremental.. in that case, maybe run fsstress for a while in a temp dir, back it up, then run it again, repeat a few times. Something along those lines.

I think backuppc is all command-line based, so it should be easy to script. Also, if we're lucky, there is a debian package for it that we can have ceph-qa-chef.git/.../default.rb install for us on the qa machines.