We are standing up a small Nautilus cluster using VMs on an old storage node, and will go through the upgrade on this test environment, just to make sure. We did have a nasty experience with our last upgrade from Luminous to Nautilus. We had installed 2 Mimic MDS daemons, with the intention of testing CephFS, but never got further. When we did the Luminous to Nautilus upgrade, the Monitors all got corrupted, and had to be rebuilt from the OSDs. Everything looked fine, until the last monitor was upgraded, then they all died. The root cause turned out to be the presence of the Mimic MDS daemons, even though they weren't ever used (beyond having them running). After we removed the Mimic MDS servers, the upgrade proceeded normally. To stop that happening again, we pinned the Ceph repo to a local server, but didn't bank on having the OS upgraded on us, and getting a newer Ceph version that way. _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |