Thank you very much Anthony and Eugen , I followed your instructions and now it works fine class are hdd and ssd , and also now we have 60 OSDS from 48 Thanks again Michel On Mon, 9 Jan 2023, 17:00 Anthony D'Atri, <anthony.datri@xxxxxxxxx> wrote: > For anyone finding this thread down the road: I wrote to the poster > yesterday with the same observation. Browsing the ceph-ansible docs and > code, to get them to deploy as they want, one may pre-create LVs and > enumerate them as explicit data devices. Their configuration also enables > primary affinity, so I suspect they’re trying the [brittle] trick of mixing > HDD and SSD OSDs in the same pool, with the SSDs forced primary for reads. > > > > > Hi, > > > > it appears that the SSDs were used as db devices (/dev/sd[efgh]). > According to [1] (I don't use ansible) the simple case is that: > > > >> [...] most of the decisions on how devices are configured to provision > an OSD are made by the Ceph tooling (ceph-volume lvm batch in this case). > > > > And I assume that this exactly what happened, ceph-volume batch deployed > the SSDs as rocksDB, not sure how to prevent ansible from doing that, but > there are probably several threads out there that explain it. > > > > Regards, > > Eugen > > > > [1] > https://docs.ceph.com/projects/ceph-ansible/en/latest/osds/scenarios.html > _______________________________________________ > ceph-users mailing list -- ceph-users@xxxxxxx > To unsubscribe send an email to ceph-users-leave@xxxxxxx > _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |