From 804ea3abc4265c035d2b3941400c842e6d6fdb96 Mon Sep 17 00:00:00 2001
From: Hongchao Zhang <hongchao.zhang@intel.com>
Date: Thu, 18 Jan 2018 16:58:42 +0800
Subject: [PATCH] LU-10052 test: relate fs_log_size to recordsize

If the backend filesystem is ZFS, the block usage difference is
related to the recordsize of it, the maximum difference are 2 blocks.
This affects several different tests that have intermittent failures.

    replay-dual test_14b, replay-single test_20b, test_89

Change-Id: I36b184587306bd2b9221e5771bf1adfe071653ca
Signed-off-by: Hongchao Zhang <hongchao.zhang@intel.com>
Reviewed-on: https://review.whamcloud.com/30916
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Nathaniel Clark <nathaniel.l.clark@intel.com>
Reviewed-by: James Nunez <james.a.nunez@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
---
 lustre/tests/replay-dual.sh    | 14 +++++++-------
 lustre/tests/replay-single.sh  | 26 ++++++++++++++++++++------
 lustre/tests/test-framework.sh |  7 ++++---
 3 files changed, 31 insertions(+), 16 deletions(-)

diff --git a/lustre/tests/replay-dual.sh b/lustre/tests/replay-dual.sh
index 7f311a2..527d0b4 100755
--- a/lustre/tests/replay-dual.sh
+++ b/lustre/tests/replay-dual.sh
@@ -2,8 +2,8 @@
 
 set -e
 
-# bug number:  LU-2012 10124 LU-8333
-ALWAYS_EXCEPT="14b     15c   21b     $REPLAY_DUAL_EXCEPT"
+# bug number:  10124 LU-8333
+ALWAYS_EXCEPT="15c   21b     $REPLAY_DUAL_EXCEPT"
 
 SAVE_PWD=$PWD
 PTLDEBUG=${PTLDEBUG:--1}
@@ -327,7 +327,7 @@ test_14b() {
 	wait_mds_ost_sync
 	wait_delete_completed
 
-	local BEFOREUSED=$(df -P $DIR | tail -1 | awk '{ print $3 }')
+	local beforeused=$(df -P $DIR | tail -1 | awk '{ print $3 }')
 
 	mkdir -p $MOUNT1/$tdir
 	$SETSTRIPE -i 0 $MOUNT1/$tdir
@@ -352,11 +352,11 @@ test_14b() {
 	wait_mds_ost_sync || error "wait_mds_ost_sync failed"
 	wait_delete_completed || error "wait_delete_complete failed"
 
-	local AFTERUSED=$(df -P $DIR | tail -1 | awk '{ print $3 }')
-	log "before $BEFOREUSED, after $AFTERUSED"
+	local afterused=$(df -P $DIR | tail -1 | awk '{ print $3 }')
+	log "before $beforeused, after $afterused"
 	# leave some margin for some files/dirs to be modified (OI, llog, etc)
-	[ $AFTERUSED -gt $((BEFOREUSED + 128)) ] &&
-		error "after $AFTERUSED > before $BEFOREUSED" || true
+	[ $afterused -le $((beforeused + $(fs_log_size))) ] ||
+		error "after $afterused > before $beforeused"
 }
 run_test 14b "delete ost orphans if gap occured in objids due to VBR"
 
diff --git a/lustre/tests/replay-single.sh b/lustre/tests/replay-single.sh
index 7821118..3327962 100755
--- a/lustre/tests/replay-single.sh
+++ b/lustre/tests/replay-single.sh
@@ -448,10 +448,14 @@ run_test 20a "|X| open(O_CREAT), unlink, replay, close (test mds_cleanup_orphans
 
 test_20b() { # bug 10480
 	local wait_timeout=$((TIMEOUT * 4))
-	local beforeused
-	local afterused
+	local extra=$(fs_log_size)
+	local n_attempts=1
+
+	sync_all_data
+	$LFS setstripe -i 0 -c 1 $DIR
+
+	local beforeused=$(df -P $DIR | tail -1 | awk '{ print $3 }')
 
-	beforeused=$(df -P $DIR | tail -1 | awk '{ print $3 }')
 	dd if=/dev/zero of=$DIR/$tfile bs=4k count=10000 &
 	while [ ! -e $DIR/$tfile ] ; do
 		usleep 60			# give dd a chance to start
@@ -468,12 +472,22 @@ test_20b() { # bug 10480
 	fail $SINGLEMDS				# start orphan recovery
 	wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
 	wait_delete_completed $wait_timeout || error "delete did not finish"
+	sync_all_data
 
-	afterused=$(df -P $DIR | tail -1 | awk '{ print $3 }')
-	(( $afterused > $beforeused + $(fs_log_size) )) &&
-		error "after $afterused > before $beforeused" ||
+	while true; do
+		local afterused=$(df -P $DIR | tail -1 | awk '{ print $3 }')
 		log "before $beforeused, after $afterused"
+
+		(( $beforeused + $extra >= $afterused )) && break
+		n_attempts=$((n_attempts + 1))
+		[ $n_attempts -gt 3 ] &&
+			error "after $afterused > before $beforeused"
+
+		wait_zfs_commit $SINGLEMDS 5
+		sync_all_data
+	done
 }
+
 run_test 20b "write, unlink, eviction, replay (test mds_cleanup_orphans)"
 
 test_20c() { # bug 10480
diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 06910a8..85756f0f 100755
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -746,11 +746,12 @@ unload_modules() {
 
 fs_log_size() {
 	local facet=${1:-$SINGLEMDS}
-	local fstype=$(facet_fstype $facet)
 	local size=0
-	case $fstype in
+
+	case $(facet_fstype $facet) in
 		ldiskfs) size=50;; # largest seen is 44, leave some headroom
-		zfs)     size=512;; # largest seen is 512
+		zfs)     size=$(lctl get_param osc.$FSNAME-OST*.import |
+			awk '/grant_block_size:/ { print $2 * 2; exit; }');;
 	esac
 
 	echo -n $size
-- 
1.8.3.1