Whamcloud - gitweb
LU-16002 ptlrpc: reduce pinger eviction time 28/47928/10
authorAlexander Boyko <alexander.boyko@hpe.com>
Fri, 16 Sep 2022 08:00:38 +0000 (04:00 -0400)
committerOleg Drokin <green@whamcloud.com>
Sat, 15 Oct 2022 05:54:34 +0000 (05:54 +0000)
commit6bdeda7afe92d61db56367875774fa074aaac0fd
tree94e48b0fccd13ff74220f2e240a921f5780f4a73
parentf792297212387c2ff1b1dd6ead1e2d451a375ade
LU-16002 ptlrpc: reduce pinger eviction time

On a server side eviction is based on PING_INTERVAL. A client
should be evicted after PING_EVICT_TIMEOUT. But eviction logic
adds additional 3 PING_INTERVAL for it. For a configuration
with obd_timeout equal to 300, addition is 225 seconds.
The second level timeout is needed when network is down for
some time. And it prevents clients evictions after first
connection.
Patch adds additional logic to check if an import is active,
and evict client faster without second level. It reduces an
eviction timeout to a PING_EVICT_TIMEOUT.

replay_dual test_0a  is based on a client eviction during recovery,
lfs df check could fail because of eviction. So complete check
similar to recovery-small.sh

Test-Parameters: testlist=recovery-small env=RECOVERY_SMALL_EXCEPT=144 serverversion=2.14
HPE-bug-id: LUS-11054
Signed-off-by: Alexander Boyko <alexander.boyko@hpe.com>
Change-Id: I4d60046ef4737f9cf95a16ac0ab63a36859b8adc
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/47928
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Alexander Zarochentsev <alexander.zarochentsev@hpe.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
lustre/ptlrpc/service.c
lustre/tests/recovery-small.sh
lustre/tests/replay-dual.sh