Open ironhalik opened 1 month ago
I've seen this "File descriptor in bad state" issue pop up several times also, it causes files not to heal. There's another bug report somewhere over here mentioning it. A temporary solution is to restart the volume (gluster volume volname stop && gluster volume volname start).
Unfortunately, restarting the volume doesn't help in any way.
Description of problem: Scenario: ~3TB and 10 million files gluster volume hosted on three nodes with SSDs, gbit ethernet, 4 cores, 16gb ram Thinly provisioned LVM. I do a snapshot of the volume:
mount the snapshot on one of the nodes:
and finally run a borg backup on the snapshot:
The full output of the command that failed:
During creation of the backup (intense read operation), I can see in logs:
The issues start after ~10 minutes of read operations. Interestingly, if I modify the volume with:
it does fail as well, but after about an hour. So seems to be more resilient.
Expected results: All read operations successful. Backup done.
- The operating system / glusterfs version: Ubuntu 24.04 Gluster 11.1 Recently upgraded from 10.5 which worked fine.