Hey all, I recently had two OSDs fail. The first one I just removed and expected replication to fix for me. Replication froze and I restarted the OSDs after seeing heartbeat failures. It allowed replication to resume but one of the OSD's RocksDB became corrupted, showing this error when I try to bring it up: rocksdb: verify_sharding unable to list column families: Corruption: CURRENT file does not end with newline bluestore(/var/lib/ceph/osd/ceph-1) _open_db erroring opening db: osd.1 0 OSD:init: unable to mount object store ** ERROR: osd init failed: (5) Input/output error I found a relevant bug report[1] with a workaround, but it requires ceph-bluestore-tool to have bluefs-import, which was only added in 16.2.11, I'm on 16.2.9 right now. Is it safe to do a patch upgrade from 16.2.9 to 16.2.11 without having an OSD running? Or an alternative way of running the tool without upgrading? Is the tool even safe to run on data from 16.2.9? Thanks, Jess [1] https://tracker.ceph.com/issues/47330 _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |