Hi, I needed to add more spinning HDD to my nodes ( SuperMicro SSG-641E-E1CR36L) and made the mistake of NOT setting up osd_auto_discovery to "false" so ceph created OSDs on all 5 new spinning HDD This was an issue as I wanted to configure the OSDs the same as existing ones ( i.e. with WAL/DB on NVME) when the other 37 drives arrive No big harm done though because I could zap them and reconfigure ( after running ceph orch apply osd --all-available-devices --unmaged=true) when I will receive the remaining 37 drives ( as I will add 6 drives on each server) The interesting part is that, for whatever reason, one of the existing SSD based OSD is down now because the SSD drive it used changed from /dev/sdp to /dev/sdu therefore there is no "block" entry under /var/lib/ceph/FSID/osd.XX I am not sure why adding spinning disks mess up the order/naming of sdd disks I would appreciate some advice regarding the best course of action to reconfigure the OSD that is down The cluster is healthy , not busy, with all the other OSDs working as expected Has 7 hosts with 12 SSD and 6 HDD drives each ( one of them is the one with issues), 2 EC 4+2 pool , 2 MDS and few metadata pools replicated on NVME There are also 3 NVME disks on each dedicated to pools Many thanks Steven _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |