Hi, I'm still unable to get our filesystem back. I now have this: fs_cluster - 0 clients ========== RANK STATE MDS ACTIVITY DNS INOS DIRS CAPS 0 rejoin cephmd4b 90.0k 89.4k 14.7k 0 1 rejoin cephmd6b 105k 105k 21.3k 0 2 failed POOL TYPE USED AVAIL fs_cluster_meta metadata 288G 55.2T fs_cluster_data data 421T 55.2T Still cannot get rid of the 3rd failed rank. But the other two currently stay in state rejoin forever. After all clients were stopped, the log complains about a 'dup inode': 2024-05-30T07:59:46.252+0200 7f2fe9146700 -1 log_channel(cluster) log [ERR] : loaded dup inode 0x1001710ea1d [12bc6a,head] v1432525092 at /homes/YYY/ZZZ/.bash_history-21032.tmp, but inode 0x1001710ea1d.head v1432525109 already exists at /homes/YYY/ZZZ/.bash_history Questions: - Is there a way to scan/repair the metadata without any MD in 'active' state ? - Is there a way to remove (or otherwise fix) the inode in question given the above inode number ? - Is the state 'rejoin' due to the inode error or because of that 3rd rank ? Regard, N.