X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Frecovery-small.sh;h=c5166e917c234a2af1b2e00ac2f9278289afad3c;hp=891ac23294c38fb48d770bcbee429b749ee99448;hb=49acd1450fff0f710924b008ab85fdb3f5f68015;hpb=fba6abdb9818b01d02ac7663e4ac9881258c8ead

diff --git a/lustre/tests/recovery-small.sh b/lustre/tests/recovery-small.sh
index 891ac23..c5166e9 100755
--- a/lustre/tests/recovery-small.sh
+++ b/lustre/tests/recovery-small.sh
@@ -2,27 +2,18 @@
 
 set -e
 
-export MULTIOP=${MULTIOP:-multiop}
 PTLDEBUG=${PTLDEBUG:--1}
-LUSTRE=${LUSTRE:-`dirname $0`/..}
+LUSTRE=${LUSTRE:-$(dirname $0)/..}
 . $LUSTRE/tests/test-framework.sh
 init_test_env $@
-. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
 init_logging
 
 ALWAYS_EXCEPT="$RECOVERY_SMALL_EXCEPT "
-# bug number for skipped test:
-ALWAYS_EXCEPT+="               "
-# UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
-
-require_dsh_mds || exit 0
-
-# also long tests: 19, 21a, 21e, 21f, 23, 27
-
-[ "$SLOW" = "no" ] && EXCEPT_SLOW=""
 
 build_test_filter
 
+require_dsh_mds || exit 0
+
 # Allow us to override the setup if we already have a mounted system by
 # setting SETUP=" " and CLEANUP=" "
 SETUP=${SETUP:-""}
@@ -1078,6 +1069,8 @@ test_26a() {      # was test_26 bug 5921 - evict dead exports by pinger
 
 	check_timeout || return 1
 
+	# make sure all imports are connected and not IDLE
+	do_facet client lfs df > /dev/null
 # OBD_FAIL_PTLRPC_DROP_RPC 0x505
 	do_facet client lctl set_param fail_loc=0x505
 	local before=$(date +%s)
@@ -1088,7 +1081,7 @@ test_26a() {      # was test_26 bug 5921 - evict dead exports by pinger
 	# the loser might have to wait for the next ping.
 	sleep $((TIMEOUT * 2 + TIMEOUT * 3 / 4))
 	do_facet client lctl set_param fail_loc=0x0
-	do_facet client df > /dev/null
+	do_facet client lfs df > /dev/null
 
 	local oscs=$(lctl dl | awk '/-osc-/ {print $4}')
 	check_clients_evicted $before ${oscs[@]}
@@ -1126,10 +1119,9 @@ test_26b() {      # bug 10140 - evict dead exports by pinger
 	# = 9 * PING_INTERVAL + PING_INTERVAL
 	# = 10 PING_INTERVAL = 10 obd_timeout / 4 = 2.5 obd_timeout
 	# let's wait $((TIMEOUT * 3)) # bug 19887
-	local rc=0
-	wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 3)) || \
-		error "Client was not evicted by ost" rc=1
-	wait_client_evicted $SINGLEMDS $MDS_NEXP $((TIMEOUT * 3)) || \
+	wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 3)) ||
+		error "Client was not evicted by ost"
+	wait_client_evicted $SINGLEMDS $MDS_NEXP $((TIMEOUT * 3)) ||
 		error "Client was not evicted by mds"
 }
 run_test 26b "evict dead exports"
@@ -1233,7 +1225,7 @@ test_51() {
         for i in $SEQ
         do
 		#echo failover in $i sec
-		log "test_$testnum: failover in $i sec"
+		log "$TESTNAME: failover in $i sec"
 		sleep $i
 		facet_failover $SINGLEMDS
         done
@@ -1534,8 +1526,8 @@ test_65() {
 	mount_client $DIR2
 
 	#grant lock1, export2
-	$SETSTRIPE -i -0 $DIR2/$tfile || return 1
-	$MULTIOP $DIR2/$tfile Ow  || return 2
+	$LFS setstripe -i -0 $DIR2/$tfile || error "setstripe failed"
+	$MULTIOP $DIR2/$tfile Ow  || error "multiop failed"
 
 #define OBD_FAIL_LDLM_BL_EVICT            0x31e
 	do_facet ost $LCTL set_param fail_loc=0x31e
@@ -1932,7 +1924,7 @@ test_105()
 	# Since the client just mounted, its last_rcvd entry is not on disk.
 	# Send an RPC so exp_need_sync forces last_rcvd to commit this export
 	# so the client can reconnect during OST recovery (LU-924, LU-1582)
-	$SETSTRIPE -i 0 $DIR/$tfile
+	$LFS setstripe -i 0 $DIR/$tfile
 	dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 conv=sync
 
         # make sure MGS's state is Partial
@@ -2034,7 +2026,7 @@ run_test 107 "drop reint reply, then restart MDT"
 
 test_108() {
 	mkdir -p $DIR/$tdir
-	$SETSTRIPE -c 1 -i 0 $DIR/$tdir
+	$LFS setstripe -c 1 -i 0 $DIR/$tdir
 
 	dd if=/dev/zero of=$DIR/$tdir/$tfile bs=1M count=256 &
 	local dd_pid=$!
@@ -2097,7 +2089,7 @@ test_110c () {
 	drop_update_reply $mdtidx "$LFS mkdir -i $mdtidx -c2 $remote_dir" ||
 		error "lfs mkdir failed"
 
-	diridx=$($GETSTRIPE -m $remote_dir)
+	diridx=$($LFS getstripe -m $remote_dir)
 	[ $diridx -eq $mdtidx ] || error "$diridx != $mdtidx"
 
 	rm -rf $DIR/$tdir || error "rmdir failed"
@@ -2266,13 +2258,15 @@ test_110k() {
 
 #define OBD_FAIL_FLD_QUERY_REQ 0x1103
 	do_facet mds2 lctl set_param fail_loc=0x1103
-	start mds2 $(mdsdevname 2) -o abort_recovery ||
+	local OPTS="$MDS_MOUNT_OPTS -o abort_recovery"
+	start mds2 $(mdsdevname 2) $OPTS ||
 		error "start MDS with abort_recovery should succeed"
 	do_facet mds2 lctl set_param fail_loc=0
 
 	# cleanup
 	stop mds2 || error "cleanup: stop mds2 failed"
-	start mds2 $(mdsdevname 2) || error "cleanup: start mds2 failed"
+	start mds2 $(mdsdevname 2) $MDS_MOUNT_OPTS ||
+		error "cleanup: start mds2 failed"
 	zconf_mount $(hostname) $MOUNT || error "cleanup: mount failed"
 	client_up || error "post-failover df failed"
 }
@@ -2287,10 +2281,10 @@ test_111 ()
 #define OBD_FAIL_MDS_CHANGELOG_INIT 0x151
 	do_facet $SINGLEMDS lctl set_param fail_loc=0x151
 	stop $SINGLEMDS || error "stop MDS failed"
-	start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) &&
+	start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS &&
 		error "start MDS should fail"
 	do_facet $SINGLEMDS lctl set_param fail_loc=0
-	start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) ||
+	start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS ||
 		error "start MDS failed"
 }
 run_test 111 "mdd setup fail should not cause umount oops"
@@ -2672,7 +2666,7 @@ test_132() {
 
 	rm -f $DIR/$tfile
 	# get a lock on client so that export would reach the stale list
-	$SETSTRIPE -i 0 $DIR/$tfile || error "setstripe failed"
+	$LFS setstripe -i 0 $DIR/$tfile || error "setstripe failed"
 	dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=fsync ||
 		error "dd failed"
 
@@ -2704,7 +2698,7 @@ test_131() {
 
 	rm -f $DIR/$tfile
 	# get a lock on client so that export would reach the stale list
-	$SETSTRIPE -i 0 $DIR/$tfile || error "setstripe failed"
+	$LFS setstripe -i 0 $DIR/$tfile || error "setstripe failed"
 	dd if=/dev/zero of=$DIR/$tfile count=1 || error "dd failed"
 
 	# another IO under the same lock
@@ -2876,6 +2870,290 @@ test_137() {
 }
 run_test 137 "late resend must be skipped if already applied"
 
+test_138() {
+	remote_mds_nodsh && skip "remote MDS with nodsh"
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
+	[[ "$MDS1_VERSION" -ge $(version_code 2.12.59) ]] ||
+		skip "Need server version newer than 2.12.59"
+
+	zconf_umount_clients $CLIENTS $MOUNT
+
+#define OBD_FAIL_TGT_RECOVERY_CONNECT 0x724
+	#delay a first step of recovey when MDS waiting clients
+	#and failing to get osp logs
+	do_facet $SINGLEMDS $LCTL set_param fail_loc=0x724 fail_val=5
+
+	facet_failover $SINGLEMDS
+
+	#waiting failover and recovery timer
+	#the valuse is based on target_recovery_overseer() wait_event timeout
+	sleep 55
+	stop $SINGLEMDS || error "stop MDS failed"
+	do_facet $SINGLEMDS $LCTL set_param fail_loc=0
+	start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS ||
+		error "start MDS failed"
+	zconf_mount_clients $CLIENTS $MOUNT
+}
+run_test 138 "Umount MDT during recovery"
+
+test_139() {
+	[ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
+	[ $MDS1_VERSION -lt $(version_code 2.13.50) ] &&
+		skip "Need MDS version at least 2.13.50"
+
+	mdt_dev=$(mdsdevname 1)
+
+	stop $SINGLEMDS || error "stop $SINGLEMDS failed"
+
+#define OBD_FAIL_OSP_INVALID_LOGID		0x2106
+	do_facet $SINGLEMDS $LCTL set_param fail_val=0x68 fail_loc=0x80002106
+	start $SINGLEMDS $mdt_dev $MDS_MOUNT_OPTS || error "Fail to start MDT"
+}
+run_test 139 "corrupted catid won't cause crash"
+
+test_140a() {
+	[ $MDS1_VERSION -lt $(version_code 2.12.58) ] &&
+		skip "Need MDS version at least 2.13.50"
+
+	[ "$SHARED_KEY" = true ] &&
+		skip "server local client incompatible with SSK keys installed"
+
+	slr=$(do_facet mds1 \
+		$LCTL get_param -n mdt.$FSNAME-MDT0000.local_recovery)
+	stack_trap "do_facet mds1 $LCTL set_param \
+		mdt.*.local_recovery=$slr" EXIT
+
+	# disable recovery for local clients
+	# so local clients should be marked with no_recovery flag
+	do_facet mds1 $LCTL set_param mdt.*.local_recovery=0
+	mount_mds_client
+
+	local cnt
+	cnt=$(do_facet mds1 $LCTL get_param "mdt.*MDT0000.exports.*.export" |
+		grep export_flags.*no_recovery | wc -l)
+	echo "$cnt clients with recovery disabled"
+	umount_mds_client
+	[ $cnt -eq 0 ] && error "no clients with recovery disabled"
+
+	# enable recovery for local clients
+	# so no local clients should be marked with no_recovery flag
+	do_facet mds1 $LCTL set_param mdt.*.local_recovery=1
+	mount_mds_client
+
+	cnt=$(do_facet mds1 $LCTL get_param "mdt.*MDT0000.exports.*.export" |
+		grep export_flags.*no_recovery | wc -l)
+	echo "$cnt clients with recovery disabled"
+	umount_mds_client
+	[ $cnt -eq 0 ] || error "$cnt clients with recovery disabled"
+}
+run_test 140a "local mount is flagged properly"
+
+test_140b() {
+	[ $MDS1_VERSION -lt $(version_code 2.12.58) ] &&
+		skip "Need MDS version at least 2.13.50"
+
+	[ "$SHARED_KEY" = true ] &&
+		skip "server local client incompatible with SSK keys installed"
+
+	slr=$(do_facet mds1 \
+		$LCTL get_param -n mdt.$FSNAME-MDT0000.local_recovery)
+	stack_trap "do_facet mds1 $LCTL set_param \
+		mdt.*.local_recovery=$slr" EXIT
+
+	# disable recovery for local clients
+	do_facet mds1 $LCTL set_param mdt.*.local_recovery=0
+
+	mount_mds_client
+	replay_barrier mds1
+	umount_mds_client
+	fail mds1
+	# Lustre: tfs-MDT0000: Recovery over after 0:03, of 2 clients 2 rec...
+	local recovery=$(do_facet mds1 dmesg |
+			 awk '/Recovery over after/ { print $6 }' | tail -1 |
+			 awk -F: '{ print $1 * 60 + $2 }')
+	(( recovery < TIMEOUT * 2 + 5 )) ||
+		error "recovery took too long $recovery > $((TIMEOUT * 2 + 5))"
+}
+run_test 140b "local mount is excluded from recovery"
+
+test_141() {
+	local oldc
+	local newc
+
+	[ $PARALLEL == "yes" ] && skip "skip parallel run"
+	combined_mgs_mds || skip "needs combined MGS/MDT"
+	( local_mode || from_build_tree ) &&
+		skip "cannot run in local mode or from build tree"
+
+	# some get_param have a bug to handle dot in param name
+	do_rpc_nodes $(facet_active_host $SINGLEMDS) cancel_lru_locks MGC
+	oldc=$(do_facet $SINGLEMDS $LCTL get_param -n \
+		'ldlm.namespaces.MGC*.lock_count')
+	fail $SINGLEMDS
+	do_rpc_nodes $(facet_active_host $SINGLEMDS) cancel_lru_locks MGC
+	newc=$(do_facet $SINGLEMDS $LCTL get_param -n \
+		'ldlm.namespaces.MGC*.lock_count')
+
+	[ $oldc -eq $newc ] || error "mgc lost locks ($oldc != $newc)"
+	return 0
+}
+run_test 141 "do not lose locks on MGS restart"
+
+test_142() {
+	[ $MDS1_VERSION -lt $(version_code 2.11.56) ] &&
+		skip "Need MDS version at least 2.11.56"
+
+	#define OBD_FAIL_MDS_ORPHAN_DELETE	0x165
+	do_facet mds1 $LCTL set_param fail_loc=0x165
+	$MULTIOP $DIR/$tfile Ouc || error "multiop failed"
+
+	stop mds1
+	start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS
+
+	wait_update_facet mds1 "pgrep orph_.*-MDD | wc -l" "0" ||
+		error "MDD orphan cleanup thread not quit"
+}
+run_test 142 "orphan name stub can be cleaned up in startup"
+
+test_143() {
+	[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.13.00) ] &&
+		skip "Need MDS version at least 2.13.00"
+	[ $PARALLEL == "yes" ] && skip "skip parallel run"
+
+	local mntpt=$(facet_mntpt $SINGLEMDS)
+	stop mds1
+	mount_fstype $SINGLEMDS || error "mount as fstype $SINGLEMDS failed"
+	do_facet $SINGLEMDS touch $mntpt/PENDING/$tfile
+	unmount_fstype $SINGLEMDS
+	start mds1 $(mdsdevname 1) $MDS_MOUNT_OPTS || error "mds1 start fail"
+
+	wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
+	wait_update_facet mds1 "pgrep orph_.*-MDD | wc -l" "0" ||
+		error "MDD orphan cleanup thread not quit"
+}
+run_test 143 "orphan cleanup thread shouldn't be blocked even delete failed"
+
+test_145() {
+	[ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs"
+	[ $(facet_active_host mds2) = $(facet_active_host mds3) ] &&
+		skip "needs mds2 and mds3 on separate nodes"
+
+	replay_barrier mds1
+
+	touch $DIR/$tfile
+
+#define OBD_FAIL_PTLRPC_DELAY_RECOV      0x507
+	echo block mds_connect from mds2
+	do_facet mds2 "$LCTL set_param fail_loc=0x507"
+
+#define OBD_FAIL_OUT_UPDATE_DROP	0x1707
+	echo block recovery updates from mds3
+	do_facet mds3 "$LCTL set_param fail_loc=0x1707"
+
+	local hard_timeout=\
+$(do_facet mds1 $LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_time_hard)
+
+	fail mds1 &
+
+	local get_soft_timeout_cmd=\
+"$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_time_soft 2>/dev/null"
+
+	echo wait until mds1 recovery_time_soft is $hard_timeout
+	wait_update $(facet_host mds1) "$get_soft_timeout_cmd" \
+"$hard_timeout" $hard_timeout
+
+	echo unblock mds_connect from mds2
+	do_facet mds2 "$LCTL set_param fail_loc=0"
+
+	echo upblock recovery updates from mds3
+	do_facet mds3 "$LCTL set_param fail_loc=0"
+
+	wait
+	[ -f $DIR/$tfile ] || error "$DIR/$tfile does not exist"
+}
+run_test 145 "connect mdtlovs and process update logs after recovery expire"
+
+test_147() {
+	local obd_timeout=200
+	local old=$($LCTL get_param -n timeout)
+	local f=$DIR/$tfile
+	local connection_count
+
+	$LFS setstripe -i 0 -c 1 $f
+	stripe_index=$($LFS getstripe -i $f)
+	if [ $stripe_index -ne 0 ]; then
+	       $LFS getstripe $f
+	       error "$f: stripe_index $stripe_index != 0" && return
+	fi
+
+	$LCTL set_param timeout=$obd_timeout
+	stack_trap "$LCTL set_param timeout=$old && client_reconnect" EXIT
+
+	# OBD_FAIL_OST_CONNECT_NET2
+	# lost reply to connect request
+	do_facet ost1 lctl set_param fail_loc=0x00000225 timeout=$obd_timeout
+	stack_trap "do_facet ost1 $LCTL set_param fail_loc=0 timeout=$old" EXIT
+
+
+	ost_evict_client
+	# force reconnect
+	$LFS df $MOUNT > /dev/null 2>&1 &
+	sleep $((obd_timeout * 3 / 4))
+
+	$LCTL get_param osc.$FSNAME-OST0000-osc-*.state
+	connection_count=$($LCTL get_param osc.$FSNAME-OST0000-osc-*.state |
+			   tac |  sed "/FULL/,$ d" | grep CONNECTING | wc -l)
+
+	echo $connection_count
+	(($connection_count >= 6)) || error "Client reconnected too slow"
+}
+run_test 147 "Check client reconnect"
+
+test_148() {
+	local wce_param="obdfilter.$FSNAME-OST0000.writethrough_cache_enable"
+	local p="$TMP/$TESTSUITE-$TESTNAME.parameters"
+	local amc=$(at_max_get client)
+	local amo=$(at_max_get ost1)
+	local timeout
+
+	at_max_set 0 client
+	at_max_set 0 ost1
+	timeout=$(request_timeout client)
+
+	[ "$(facet_fstype ost1)" = "ldiskfs" ] && {
+		# save old r/o cache settings
+		save_lustre_params ost1 $wce_param > $p
+
+		# disable r/o cache
+		do_facet ost1 "$LCTL set_param -n $wce_param=0"
+	}
+
+	$LFS setstripe -i 0 -c 1 $DIR/$tfile
+	dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 oflag=direct
+	cp $DIR/$tfile $TMP/$tfile
+	#define OBD_FAIL_OST_BRW_PAUSE_BULK2     0x227
+	do_facet ost1 $LCTL set_param fail_loc=0x80000227
+	do_facet ost1 $LCTL set_param fail_val=$((timeout+2))
+	dd if=/dev/urandom of=$DIR/$tfile bs=4096 count=1 conv=notrunc,fdatasync
+	dd if=/dev/zero of=$DIR/$tfile bs=4096 count=1 conv=notrunc,fdatasync
+	sleep 2
+	cancel_lru_locks osc
+	cmp -b $DIR/$tfile $TMP/$tfile || error "wrong data"
+
+	rm -f $DIR/$tfile $TMP/$tfile
+
+	at_max_set $amc client
+	at_max_set $amo ost1
+
+	[ "$(facet_fstype ost1)" = "ldiskfs" ] && {
+		# restore initial r/o cache settings
+		restore_lustre_params < $p
+	}
+
+	return 0
+}
+run_test 148 "data corruption through resend"
+
 complete $SECONDS
 check_and_cleanup_lustre
 exit_status