From 45569f8327d70593c2a77bf61819e299fdd1d4d5 Mon Sep 17 00:00:00 2001
From: tappro <tappro>
Date: Wed, 24 Sep 2008 07:50:18 +0000
Subject: [PATCH] - prevent the growing of recovery timeout b:16389
i:nathan, rread
---
lustre/ptlrpc/service.c | 20 +++++++++++++-------
lustre/tests/replay-dual.sh | 29 +++++++++++++++++++++++++++++
2 files changed, 42 insertions(+), 7 deletions(-)
diff --git a/lustre/ptlrpc/service.c b/lustre/ptlrpc/service.c
index 59cfe9799e..aaf891558f 100644
--- a/lustre/ptlrpc/service.c
+++ b/lustre/ptlrpc/service.c
@@ -699,15 +699,21 @@ static int ptlrpc_at_send_early_reply(struct ptlrpc_request *req,
RETURN(-ENOSYS);
}
- if (extra_time) {
- /* Fake our processing time into the future to ask the
- * clients for some extra amount of time */
- extra_time += cfs_time_current_sec() -
- req->rq_arrival_time.tv_sec;
- at_add(&svc->srv_at_estimate, extra_time);
+ if (req->rq_export && req->rq_export->exp_obd->obd_recovering) {
+ newdl = cfs_time_current_sec() +
+ req->rq_export->exp_obd->obd_recovery_timeout;
+ } else {
+ if (extra_time) {
+ /* Fake our processing time into the future to ask the
+ * clients for some extra amount of time */
+ extra_time += cfs_time_current_sec() -
+ req->rq_arrival_time.tv_sec;
+ at_add(&svc->srv_at_estimate, extra_time);
+ }
+ newdl = req->rq_arrival_time.tv_sec +
+ at_get(&svc->srv_at_estimate);
}
- newdl = req->rq_arrival_time.tv_sec + at_get(&svc->srv_at_estimate);
if (req->rq_deadline >= newdl) {
/* We're not adding any time, no need to send an early reply
(e.g. maybe at adaptive_max) */
diff --git a/lustre/tests/replay-dual.sh b/lustre/tests/replay-dual.sh
index cf397cba53..ffa2186d29 100755
--- a/lustre/tests/replay-dual.sh
+++ b/lustre/tests/replay-dual.sh
@@ -369,6 +369,35 @@ test_19() { # Bug 10991 - resend of open request does not fail assertion.
}
run_test 19 "resend of open request"
+test_20() { #16389
+ BEFORE=`date +%s`
+ replay_barrier mds
+ touch $MOUNT1/a
+ touch $MOUNT2/b
+ umount $MOUNT2
+ facet_failover mds
+ df $MOUNT1 || return 1
+ rm $MOUNT1/a
+ zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
+ TIER1=$((`date +%s` - BEFORE))
+ echo "First recovery time is $TIER1 sec"
+ BEFORE=`date +%s`
+ replay_barrier mds
+ touch $MOUNT1/a
+ touch $MOUNT2/b
+ umount $MOUNT2
+ facet_failover mds
+ df $MOUNT1 || return 1
+ rm $MOUNT1/a
+ zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
+ TIER2=$((`date +%s` - BEFORE))
+ echo "Second recovery time is $TIER2 sec"
+ [ $TIER2 -ge $((TIER1 * 2)) ] && \
+ error "recovery time is growing $TIER2 > $TIER1"
+ return 0
+}
+run_test 20 "recovery time is not increasing"
+
equals_msg `basename $0`: test complete, cleaning up
SLEEP=$((`date +%s` - $NOW))
[ $SLEEP -lt $TIMEOUT ] && sleep $SLEEP
--
GitLab