Amusingly, that's what I'm working on this week. http://tracker.ceph.com/issues/7862 There are pretty good reasons for why it works the way it does right now, but it certainly is unexpected. -Sam On Thu, Nov 6, 2014 at 7:18 AM, Chad William Seys <cwseys@xxxxxxxxxxxxxxxx> wrote: > Hi Sam, > >> Sounds like you needed osd 20. You can mark osd 20 lost. >> -Sam > > Does not work: > > # ceph osd lost 20 --yes-i-really-mean-it > osd.20 is not down or doesn't exist > > > Also, here is an interesting post which I will follow from October: > http://lists.ceph.com/pipermail/ceph-users-ceph.com/2014-October/044059.html > > " > Hello, all. I got some advice from the IRC channel (thanks bloodice!) that I > temporarily reduce the min_size of my cluster (size = 2) from 2 down to 1. > That immediately caused all of my incomplete PGs to start recovering and > everything seemed to come back OK. I was serving out and RBD from here and > xfs_repair reported no problems. So... happy ending? > > What started this all was that I was altering my CRUSH map causing significant > rebalancing on my cluster which had size = 2. During this process I lost an > OSD (osd.10) and eventually ended up with incomplete PGs. Knowing that I only > lost 1 osd I was pretty sure that I hadn't lost any data I just couldn't get > the PGs to recover without changing the min_size. > " > > It is good that this worked for him, but it also seems like a bug that it > worked! (I.e. ceph should have been able to recover on its own without weird > workarounds.) > > I'll let you know if this works for me! > > Thanks, > Chad. > _______________________________________________ > ceph-users mailing list > ceph-users@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |