[LU-9931] recovery-*-scale REQFAIL calculation defect

Duplicate — Thu, 5 Oct 2017 20:43:27 +0000

REQFAIL is the number of times that a sleep is allowed to be
less than $MINSLEEP before the test is considered a fail.
The result of

  "DURATION / SERVER_FAILOVER_PERIOD * REQFAIL_PERCENT / 100"

may not be an integer (165.6) and test fails with :

  "Failed to load with for a minimum
  period of 166 times ( REQFAIL=165 )".

The example of test failure :

==== Checking the clients loads AFTER failed client reintegrated -- failure NOT OK
WARNING: failover, client reintegration and check_client_loads time exceeded SERVER_FAILOVER_PERIOD - MINSLEEP!
Failed to load the filesystem with I/O for a minimum period of 120 166 times ( REQFAIL=165 ).
This iteration, the load was only applied for sleep=63 seconds.
Estimated max recovery time : 1475
Probably the hardware is taking excessively long time to boot.
Try to increase SERVER_FAILOVER_PERIOD (current is 300), bug 20918
2017-06-06 20:08:31 Terminating clients loads ...
Duration:               49680
Server failover period: 300 seconds
Exited after:           49810 seconds
Number of failovers before exit:
mds1 failed over 166 times
Status: FAIL: rc=6

Whamcloud Community JIRA

[LU-9931] recovery-*-scale REQFAIL calculation defect