Details
-
Improvement
-
Resolution: Fixed
-
Minor
-
None
-
None
-
4771
Description
In current handle_recovery_req, when setting the timeout for waiting next replay
to = lustre_msg_get_timeout(req->rq_reqmsg);
extend_recovery_timer(class_exp2obd(req->rq_export), to);
It did not consider the netlatency and early reply might extend the req timeout on client side, which might cause some replay tests failed like replay-dual 9, 10.