Hi, I think the ceph-users list is more appropriate for your request, I removed the dev list from the recipients If I understand correctly, you should still have one surviving monitor, is that correct? The quickest way to restore the quorum would probably be to edit the monmap (make a backup) by removing the dead monitors and startup the surviving monitor. This would give you a single-mon cluster, from there you could then redeploy the other monitors and restore OSD functionality.
I tried reinstalling the system on one node and recreating a monitor using the data taken from /var/lib/ceph/osd/, but I don't see the OSDs. This is the situation…
Is this a cephadm managed cluster? In that case the OSDs have their data directory in /var/lib/ceph/{CLUSTER_FSID}/osd.{OSD_ID}/. If it's not cephadm managed, you'll need to provide more information with details and logs etc. Regards, Eugen Zitat von FRANCESCO NAPOLI <francesco.napoli@cnr.it>:
Good morning, Ceph was installed on Ubuntu Server 24.04 with cephadm. The cluster consists of 4 nodes; the first one also acted as a monitor in the initial state, but as all the others were added, in addition to being OSD hosts, they also acted as monitors. These machines had the major flaw of not having the operating system disk in RAID, and after 20 days, 3 nodes failed. The OSD disks are fine... I tried reinstalling the system on one node and recreating a monitor using the data taken from /var/lib/ceph/osd/, but I don't see the OSDs. This is the situation… thanks for the support.
-- Napoli Francesco Gruppo Reti e Sistemi Informativi Istituto di Fisiologia Clinica - CNR Tel. 050.315.2344