From 7f89e8c8975fcc82983c5756438861d66e64ec23 Mon Sep 17 00:00:00 2001
From: Etienne AUJAMES <etienne.aujames@cea.fr>
Date: Mon, 27 Mar 2023 12:14:34 +0200
Subject: [PATCH] LU-16408 tests: fix replay-dual test 33

Client can be evicted in REPLAY_LOCK. Wait REPLAY_WAIT import state
before aborting the recovery on the MDS.

When unmounting a combined MDT and MGT, the imperative recovery is
disabled. So, we have to force update the client import states
(MGC/MDC).

Test-Parameters: trivial
Test-Parameters: testlist=replay-dual env=ONLY="33",ONLY_REPEAT=50
Test-Parameters: testlist=replay-dual
Test-Parameters: testlist=replay-dual
Test-Parameters: testlist=replay-dual
Test-Parameters: testlist=replay-dual
Fixes: 1a79d395dd ("LU-15935 target: keep track of multirpc slots in last_rcvd")
Signed-off-by: Etienne AUJAMES <eaujames@ddn.com>
Change-Id: I0869fe968a18795dae39cf39a7009cf444820017
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/50434
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Jian Yu <yujian@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
---
 lustre/tests/replay-dual.sh    | 22 ++++++++++++++++++++--
 lustre/tests/test-framework.sh | 23 +++++++++++++++++++++++
 2 files changed, 43 insertions(+), 2 deletions(-)

diff --git a/lustre/tests/replay-dual.sh b/lustre/tests/replay-dual.sh
index 693a299..85836e68b 100755
--- a/lustre/tests/replay-dual.sh
+++ b/lustre/tests/replay-dual.sh
@@ -1278,9 +1278,20 @@ test_33() { # LU-15935
 
 	[[ "$mds1_FSTYPE" == "ldiskfs" ]] || skip "ldiskfs only test"
 
+	local at_min_old
+	at_min_old=$(at_min_get ost1)
+	at_min_set 60 ost
+	stack_trap "at_min_set $at_min_old ost"
+
+	# Clean old locks
+	cancel_lru_locks mdc
+
 	clients_up
 	stop mds1
 
+	# No IR -> force the update of mdc state
+	! combined_mgs_mds || $LCTL get_param mdc.*.ping || true
+
 	# check for OBD_INCOMPAT_MULTI_RPCS (0x400) in last_rcvd
 	last_rcvd_check_incompat_flag mds1 0x400 ||
 		error "1st failover: OBD_INCOMPAT_MULTI_RPCS is not set on MDT0000 last_rcvd"
@@ -1289,18 +1300,25 @@ test_33() { # LU-15935
 	umount -f $MOUNT2
 
 	mount_facet mds1
-	wait_clients_import_state "$HOSTNAME" mds1 "\(REPLAY_WAIT\|REPLAY_LOCKS\)"
+	wait_clients_import_state "$HOSTNAME" mds1 "REPLAY_WAIT"
 
 	do_facet mds1 $LCTL --device $(convert_facet2label mds1) abort_recovery
 	wait_clients_import_state "$HOSTNAME" mds1 "FULL"
+	wait_recovery_complete mds1
+	sleep 5
 	stop mds1
 
 	last_rcvd_check_incompat_flag mds1 0x400 ||
 		error "2sd failover: OBD_INCOMPAT_MULTI_RPCS is not set on MDT0000 last_rcvd"
 
 	mount_facet mds1
-	zconf_mount $HOSTNAME $MOUNT2
+
+	# No IR -> force the update of mgc state
+	! combined_mgs_mds || $LCTL get_param mgc.*.ping || true
+
+	zconf_mount $HOSTNAME $MOUNT2 || error "Fail to mount $MOUNT2"
 	wait_clients_import_state "$HOSTNAME" mds1 "FULL"
+	wait_recovery_complete mds1
 }
 run_test 33 "Check for OBD_INCOMPAT_MULTI_RPCS in last_rcvd after abort_recovery"
 
diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 7abd2083..54e5f35 100755
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -6352,6 +6352,29 @@ at_max_set() {
 	do_nodes $hosts lctl set_param at_max=$at_max
 }
 
+at_min_get() {
+	at_get $1 at_min
+}
+
+at_min_set() {
+	local at_min=$1
+	shift
+
+	local facet
+	local hosts
+
+	for facet in "$@"; do
+		if [ $facet == "ost" ]; then
+			facet=$(get_facets OST)
+		elif [ $facet == "mds" ]; then
+			facet=$(get_facets MDS)
+		fi
+		hosts=$(expand_list $hosts $(facets_hosts $facet))
+	done
+
+	do_nodes $hosts lctl set_param at_min=$at_min
+}
+
 ##################################
 # OBD_FAIL funcs
 
-- 
1.8.3.1