Finally, after having corruption with the MDS I had no choice but to try to manually repair the PG.
Following the procedure on the blog post from Ceph at http://ceph.com/planet/ceph-manually-repair-object/ I was able to get the PG back to active+clean, ceph pg repair wasn't still not working and an automatic deep-scrub confirmed the object was still faulty.On Fri, Dec 18, 2015 at 10:42 AM, Jérôme Poulin <jeromepoulin@xxxxxxxxx> wrote:
Is there a way to ask the OSD directly to deep-scrub without using the monitor? Is there a known issue about commands getting ignored?I also tried to fix the defective object myself in case it was hanging something, now the object has the same checksum on each OSD.I tried restarting the monitor in charge to elect a new one and restart each affected OSD for the inconsistent PG with no success.The only flag set is noout since this cluster is too small, but automatic deep-scrubs are working and are logged both in ceph.log and the OSD log.I verified the log of each OSD which had the PG and confirmed that nothing was logged, no repair, no deep-scrub. After trying deep-scrubbing manually other PGs, I confirmed that my requests were completely ignored.After reparing the physical issue, I went and tried ceph pg repair, no action, then I tried ceph pg deep-scrub, still no action.Good day everyone,I currently manage a Ceph cluster running Firefly 0.80.10, we had some maintenance which implied stopping OSD and starting them back again. This caused one of the hard drive to notice it had a bad sector and then Ceph to mark it as inconsistent.
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
