Hi all, Every so often ceph shows an error message putting the complete system on read-only mode. The only way to enable the system is to restart the MDS. My MDS log shows: 022-06-22T22:08:34.192250+0200 mds.mds00 (mds.0) 28 : cluster [ERR] failed to commit dir 0x10125421110 object, errno -2 2022-06-22T22:08:34.192264+0200 mds.mds00 (mds.0) 29 : cluster [WRN] force file system read-only The object should be stored in the metada pool because I can list this object executing: [root@osd00 ~]# rados --pool metadata ls | grep 10125421110 10125421110.00000000 The object failed is different every time the error is shown but I can't find the path to this directory in order to check if it exists. I'm updating the complete production system from OCTOPUS to PACIFIC versions, now my 3 MDSs are in pacific version but the OSDs continue in OCTOPUS version. MDS,MON: ceph version 16.2.7 OSD: ceph version 15.2.4 Is there any way to fix the metadata failed objects ? Thanks in advance. _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |