[LU-9800] recovery-mds-scale test_failover_mds: test_failover_mds returned 1

• Type: Bug
• Resolution: Unresolved
• Priority: Minor
• Affects Version/s: Lustre 2.10.1, Lustre 2.11.0, Lustre 2.10.4
• Environment: trevis, failover; server: EL7, zfs, branch master, v2.10.50.3, b3612 ; client: EL7, branch master, v2.10.50.3, b3612
• Severity: 3
• Assignee: WC Triage
• Reporter: James Casper

Description

https://testing.hpdd.intel.com/test_sessions/55045e59-2766-4676-91ae-45a2fa2f4e91

The dd client loads for test_failover_mds and test_failover_ost are running out of space.

This looks different than LU-5788 because the recovery-mds subtests ran for 22 hours. The
LU-5788 failures normally happened in less than a minute.

From both test_logs:

Client load failed 

From both run_dd_debug logs:

dd: error writing ‘/mnt/lustre/d0.dd-trevis-49vm5.trevis.hpdd.intel.com/dd-file’: No space left on device
964211+0 records in
964210+0 records out
+ '[' 1 -eq 0 ']'
2017-07-11 23:55:13: dd failed