[LU-12066] recovery-small test 26b fails with “Client was not evicted by ost rc=1”

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Affects Version/s: Lustre 2.13.0, Lustre 2.10.7, Lustre 2.14.0, Lustre 2.12.4, Lustre 2.15.3
• Labels: failover
• Environment: failover test session
• Severity: 3
• Assignee: Hongchao Zhang
• Reporter: James Nunez (Inactive)

Description

recovery-small test_26b fails with “Client was not evicted by ost rc=1”. We only see this issue with failover testing

Looking at a recent failure at https://testing.whamcloud.com/test_sets/e21fbf5e-4500-11e9-9720-52540065bddc , in the suite_log, we see an error at the beginning of the test

== recovery-small test 26b: evict dead exports ======================================================= 09:09:58 (1552381798)
CMD: trevis-42vm12 lctl get_param -n timeout
trevis-42vm1: error: invalid path '/mnt/lustre': Input/output error
Starting client: trevis-42vm1.trevis.whamcloud.com:  -o user_xattr,flock trevis-42vm11:trevis-42vm12:/lustre /mnt/lustre2
CMD: trevis-42vm1.trevis.whamcloud.com mkdir -p /mnt/lustre2
CMD: trevis-42vm1.trevis.whamcloud.com mount -t lustre -o user_xattr,flock trevis-42vm11:trevis-42vm12:/lustre /mnt/lustre2
CMD: trevis-42vm12 lctl get_param -n mdt.lustre-MDT0000.num_exports
CMD: trevis-42vm5 lctl get_param -n obdfilter.lustre-OST0000.num_exports
starting with 4 OST and 12 MDS exports
…
CMD: trevis-42vm5 lctl get_param -n *.lustre-OST0000.num_exports | cut -d' ' -f2
Update not seen after 60s: wanted '3' got '4'
 recovery-small test_26b: @@@@@@ FAIL: Client was not evicted by ost rc=1 

On client 1 (vm1) we see an error in the console logs

======================================================= 09:09:58 \(1552381798\)
[261204.673102] Lustre: DEBUG MARKER: == recovery-small test 26b: evict dead exports ======================================================= 09:09:58 (1552381798)
[261205.856521] Lustre: Evicted from MGS (at 10.9.3.160@tcp) after server handle changed from 0x2ef44212d062fce8 to 0x2ef44212d0630260
[261210.854799] LustreError: 10265:0:(file.c:3644:ll_inode_revalidate_fini()) lustre: revalidate FID [0x200000007:0x1:0x0] error: rc = -5
…

There’s nothing obviously wrong looking at the rest of the console logs.

Log for more failures are at:
2.10.7 RC1 - https://testing.whamcloud.com/test_sets/88e90b9e-429d-11e9-a256-52540065bddc
2.10.6.79 - https://testing.whamcloud.com/test_sets/2d788f9c-41a2-11e9-b98a-52540065bddc
2.10.6.79 - https://testing.whamcloud.com/test_sets/1b0bebc4-4143-11e9-92fe-52540065bddc
2.10.6.63 - https://testing.whamcloud.com/test_sets/58d6004e-3cce-11e9-8e92-52540065bddc
2.10.6.62 - https://testing.whamcloud.com/test_sets/1b8704d2-39da-11e9-8f69-52540065bddc

We see a similar failure with master failover testing, but the logs do not have the ll_inode_revalidate_fini() error in the client console log and don’t have the ‘invalid path’ error. In some cases, recovery-small test 26a fails before the 26b failure.
2.12.51.105 - https://testing.whamcloud.com/test_sets/3bae48de-4456-11e9-9720-52540065bddc
2.12.51.98 - https://testing.whamcloud.com/test_sets/f7eacb0a-41fc-11e9-a256-52540065bddc
2.12.51.98 - https://testing.whamcloud.com/test_sets/bfd286aa-4160-11e9-8e92-52540065bddc
2.12.51.97 - https://testing.whamcloud.com/test_sets/4e759c4a-3f3a-11e9-a256-52540065bddc
2.12.51.79 - https://testing.whamcloud.com/test_sets/d624519c-3f1f-11e9-92fe-52540065bddc
2.12.51.97 - https://testing.whamcloud.com/test_sets/6cda3f0e-3e88-11e9-92fe-52540065bddc

Issue Links

• is related to: LU-14947 recovery-small test_26a: client not evicted from OST
• is related to: LU-10632 recovery-small test 26a fails with ‘client not evicted from OST’
• is related to: LU-4172 Test failure on test suite recovery-small test_26b: Client was not evicted by ost rc=1