At Mon, 5 Aug 2013 17:15:51 +0800, Liu Yuan wrote: > > On Mon, Aug 05, 2013 at 10:53:04AM +0200, Valerio Pachera wrote: > > 2013/8/5 Liu Yuan <namei.unix at gmail.com>: > > >> vdi check had nothing to fix. > > > > From sheep.log I see it has done some a recover right after the > > cluster has been restarted. > > > > Aug 05 09:24:42 [main] main(878) shutdown > > Aug 05 09:28:52 [main] md_add_disk(142) /mnt/sheep/dsk02, nr 1 > > Aug 05 09:28:52 [main] send_join_request(832) IPv4 ip:192.168.2.44 port:7000 > > Aug 05 09:28:52 [main] for_each_object_in_stale(384) /mnt/sheep/dsk02/.stale > > Aug 05 09:28:52 [main] main(874) sheepdog daemon (version > > 0.6.0_179_gfba98cd) started > > Aug 05 09:28:53 [main] recover_object_main(632) done:1 count:1549, > > oid:9533ed000000e6 > > Aug 05 09:28:53 [main] recover_object_main(632) done:2 count:1549, > > oid:19093f00000400 > > Aug 05 09:28:53 [main] recover_object_main(632) done:3 count:1549, > > oid:9533ed0000023a > > <cut> > > Seems that our restart of cluster is broken: an unexpected recovery is triggered > No node refuses to restart? If a cluster is shutdown, the cluster is in a > wait-for-join state untill all the nodes come back and shouldn't expect any > recovery. Just FYI, the log output is correct. The current master triggers object recovery even if sheepdog is cleanly stopped. See the commit 9b2dc24. Thanks, Kazutaka |