I would set min_size back to 2 for general running, but put it down to 1 during planned maintenance. There are a lot of threads on the ML talking about why you shouldn't run with min_size of 1.
On Thu, Aug 10, 2017, 11:36 PM Hyun Ha <hfamil...@gmail.com> wrote: > Thanks for reply. > > In my case, it was an issue about min_size of pool. > > # ceph osd pool ls detail > pool 5 'volumes' replicated size 2 min_size 2 crush_ruleset 0 object_hash > rjenkins pg_num 512 pgp_num 512 last_change 844 flags hashpspool > stripe_width 0 > removed_snaps [1~23] > > when replicated size=2 and min_size=2 is set, and osd goes down, ceph > cluster go into Err state and client I/O goes hang. > > ceph status log> > health HEALTH_ERR > 310 pgs are stuck inactive for more than 300 seconds > 35 pgs backfill_wait > 3 pgs backfilling > 38 pgs degraded > 382 pgs peering > 310 pgs stuck inactive > 310 pgs stuck unclean > 39 pgs undersized > 263 requests are blocked > 32 sec > > you can simply reproduce that. > so I solved this by set min_size=1 using "ceph osd pool set volumes > min_size 1" command. > It is very strange thing because if min_size can occurs big problem to > ceph cluster, ceph would not allow to set same value with replicated_size. > > Thanks. > > 2017-08-10 23:33 GMT+09:00 David Turner <drakonst...@gmail.com>: > >> When the node remote, are the osds being marked down immediately? If the >> node were to reboot, but not Mark the osds down, then all requires to those >> osds would block until they got marked down. >> >> On Thu, Aug 10, 2017, 5:46 AM Hyun Ha <hfamil...@gmail.com> wrote: >> >>> Hi, Ramirez >>> >>> I have exactly same problem as yours. >>> Did you solved that issue? >>> Do you have expireences or solutions? >>> >>> Thank you. >>> _______________________________________________ >>> ceph-users mailing list >>> email@example.com >>> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >>> >> >
_______________________________________________ ceph-users mailing list firstname.lastname@example.org http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com