[LU-16073] double lctl snapshot_mount results in unmount

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Affects Version/s: Upstream, Lustre 2.15.0
• Labels: ZFS server
• Environment: Lustre filesystem with ZFS as backend filesystem.

Epic/Theme

zfs

Epic

zfs

• Assignee: Akash B
• Reporter: Akash B

Description

HPE bug-id: LUS-10650
 
When snapshot fs is mounted on the server already and if < lctl snapshot_mount > is issued again, snapshot fs is getting unmounted from the server.

[root@cslmo4702 ~]# mount -t lustre
pool-mds65/mdt65 on /data/mdt65 type lustre (ro,svname=testfs-MDT0000,mgs,osd=osd-zfs)
pool-mds65/mdt65@snap_test_fo on /mnt/snap_test_fo_MDT0000 type lustre (ro,svname=36cd4520-MDT0000,nomgs,rdonly_dev,mgs,osd=osd-zfs)

[root@cslmo4702 ~]# lctl snapshot_list -F testfs
filesystem_name: testfs
snapshot_name: snap_test_fo
snapshot_fsname: 36cd4520
modify_time: Tue Dec 7 13:35:05 2021
create_time: Tue Dec 7 13:35:05 2021
status: mounted

[root@cslmo4702 ~]# lctl snapshot_mount -F testfs -n snap_test_fo
Can't mount the snapshot snap_test_fo: No such process

[root@cslmo4702 ~]# lctl snapshot_list -F testfs
filesystem_name: testfs
snapshot_name: snap_test_fo
snapshot_fsname: 36cd4520
modify_time: Tue Dec 7 13:35:05 2021
create_time: Tue Dec 7 13:35:05 2021
status: not mount

[root@cslmo4702 ~]# mount -t lustre
pool-mds65/mdt65 on /data/mdt65 type lustre (ro,svname=testfs-MDT0000,mgs,osd=osd-zfs)

If we try the same operation with the `mount -t lustre` command instead of `lctl snapshot_mount` command then it works correctly (EALREADY is returned).

[root@cslmo4702 ~]# mount -t lustre -o rdonly_dev pool-mds65/mdt65@snap_test_fo /mnt/snap_test_fo_MDT0000
mount.lustre: mount pool-mds65/mdt65@snap_test_fo at /mnt/snap_test_fo_MDT0000 failed: Operation already in progress
The target service is already running. (pool-mds65/mdt65@snap_test_fo)