I had a hard reset on a ceph node, and one of the OSDs is not starting due to leveldb error. At that moment, the node was trying to start up, but there was no actual writing of new data: 2016-01-27 12:00:37.068431 7f367f654880 0 ceph version 0.94.5 (9764da52395923e0b32908d83a9f7304401fee43), process ceph-osd, pid 24734 2016-01-27 12:00:37.115800 7f367f654880 0 filestore(/var/lib/ceph/osd/ceph-26) backend xfs (magic 0x58465342) 2016-01-27 12:00:37.133031 7f367f654880 0 genericfilestorebackend(/var/lib/ceph/osd/ceph-26) detect_features: FIEMAP ioctl is supported and appears to work 2016-01-27 12:00:37.133042 7f367f654880 0 genericfilestorebackend(/var/lib/ceph/osd/ceph-26) detect_features: FIEMAP ioctl is disabled via 'filestore fiemap' config option 2016-01-27 12:00:37.136538 7f367f654880 0 genericfilestorebackend(/var/lib/ceph/osd/ceph-26) detect_features: syncfs(2) syscall fully supported (by glibc and kernel) 2016-01-27 12:00:37.137584 7f367f654880 0 xfsfilestorebackend(/var/lib/ceph/osd/ceph-26) detect_feature: extsize is supported and kernel 3.10.0-123.el7.x86_64 >= 3.5 2016-01-27 12:00:37.176226 7f367f654880 -1 filestore(/var/lib/ceph/osd/ceph-26) Error initializing leveldb : Corruption: 1 missing files; e.g.: /var/lib/ceph/osd/ceph-26/current/omap/075074.sst 2016-01-27 12:00:37.176286 7f367f654880 -1 osd.26 0 OSD:init: unable to mount object store 2016-01-27 12:00:37.176315 7f367f654880 -1 ** ERROR: osd init failed: (1) Operation not permitted The file 075074.sst is missing indeed. Since I was not able to restart the OSD, and I could not find information on recovering the leveldb, I marked the OSD as lost, but then I got 3 incomplete OSDs. I tried to follow the recovery howto at https://ceph.com/community/incomplete-pgs-oh-my/, but it stepped on the same error on leveldb, having the same dependency. Is there any means to recover from this situation? To check and recover leveldb as good as possible? Or, alternatively, to get rid of the incomplete status, even with the penalty of losing some objects? Regards. -- |
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com