A quick update from me. A complete uninstall (following these instructions) and reinstall and rebuild of Ceph fixed my issue, but it also appears it was exacerbated by the fact my Ceph nodes regularly had a clock skew of over 200ms. This was caused by having my Ceph nodes sync their NTP with the...
Thankfully I've managed to get everything running again from backups and a NAS sharing over NFS, and I've only lost a small amount of data which it simple to recreate. I'm going to try completely recreating the ceph pools and OSDs and migrating the data back onto them (after testing of course!).
I've tried removing and recreating the OSDs affected, and they seem to create OK, but then fall over with the error above. As I understand it the sharding is an OSD thing, not a pool thing, so I don't understand why this would be the case. Hence asking if I can create an OSD that's not sharded...
Sadly, despite reading the docs, I've just been hit by this. I now have a number of OSDs that start and then crash with the following error. Is there way to create a non-sharded Octopus OSD in Pacific? I'm thinking that might be a way of getting my pool happy again. Can anyone help?
NOTIFY...
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.