X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Freplay-dual.sh;h=9afcb369fe2fb478be63c954599727ae7fe1c2de;hp=b9ca877dba0bd2aa20e53d0f3a96f7e14b89728a;hb=6fcbb85b3a9a618e8dd7ee14c9fc7da7d23dab5d;hpb=9f83cc191dbed9c28e58f7886e8017ab3be55d93

diff --git a/lustre/tests/replay-dual.sh b/lustre/tests/replay-dual.sh
index b9ca877..9afcb36 100755
--- a/lustre/tests/replay-dual.sh
+++ b/lustre/tests/replay-dual.sh
@@ -1,4 +1,6 @@
 #!/bin/bash
+# -*- mode: Bash; tab-width: 4; indent-tabs-mode: t; -*-
+# vim:shiftwidth=4:softtabstop=4:tabstop=4:
 
 set -e
 
@@ -13,16 +15,9 @@ CLEANUP=${CLEANUP:-""}
 MOUNT_2=${MOUNT_2:-"yes"}
 . $LUSTRE/tests/test-framework.sh
 
-if [ "$FAILURE_MODE" = "HARD" ] && mixed_ost_devs; then
-    CONFIG_EXCEPTIONS="17"
-    echo -n "Several ost services on one ost node are used with FAILURE_MODE=$FAILURE_MODE. "
-    echo "Except the tests: $CONFIG_EXCEPTIONS"
-    ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS"
-fi
-
 init_test_env $@
-
 . ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh}
+init_logging
 
 remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0
 
@@ -32,7 +27,7 @@ build_test_filter
 
 check_and_setup_lustre
 MOUNTED=$(mounted_lustre_filesystems)
-if ! $(echo $MOUNTED | grep -w -q $MOUNT2); then
+if ! $(echo $MOUNTED' ' | grep -w -q $MOUNT2' '); then
     zconf_mount $HOSTNAME $MOUNT2
     MOUNTED2=yes
 fi
@@ -42,6 +37,66 @@ rm -rf $DIR/[df][0-9]*
 
 [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
 
+# LU-482 Avert LVM and VM inability to flush caches in pre .33 kernels
+if [ $LINUX_VERSION_CODE -lt $(version_code 2.6.33) ]; then
+	sync
+	do_facet $SINGLEMDS "sync; sleep 10; sync; sleep 10; sync"
+fi
+
+LU482_FAILED=$(mktemp -u $TMP/$TESTSUITE.lu482.XXXXXX)
+test_0a() {
+	echo "Check file is LU482_FAILED=$LU482_FAILED"
+	touch $MOUNT2/$tfile-A # force sync FLD/SEQ update before barrier
+	replay_barrier $SINGLEMDS
+#define OBD_FAIL_PTLRPC_FINISH_REPLAY | OBD_FAIL_ONCE
+	touch $MOUNT2/$tfile
+	createmany -o $MOUNT1/$tfile- 50
+	$LCTL set_param fail_loc=0x80000514
+	facet_failover $SINGLEMDS
+	[ -f "$LU482_FAILED" ] && skip "LU-482 failure" && return 0
+	client_up || return 1
+	umount -f $MOUNT2
+	client_up || return 1
+	zconf_mount `hostname` $MOUNT2 || error "mount2 fais"
+	unlinkmany $MOUNT1/$tfile- 50 || return 2
+	rm $MOUNT2/$tfile || return 3
+	rm $MOUNT2/$tfile-A || return 4
+}
+run_test 0a "expired recovery with lost client"
+
+if [ -f "$LU482_FAILED" ]; then
+	log "Found check file $LU482_FAILED, aborting test script"
+	rm -vf "$LU482_FAILED"
+	complete $(basename $0) $SECONDS
+	do_nodes $CLIENTS umount -f $MOUNT2 || true
+	do_nodes $CLIENTS umount -f $MOUNT || true
+	# copied from stopall, but avoid the MDS recovery
+    for num in `seq $OSTCOUNT`; do
+        stop ost$num -f
+        rm -f $TMP/ost${num}active
+    done
+    if ! combined_mgs_mds ; then
+        stop mgs
+    fi
+
+	exit_status
+fi
+
+test_0b() {
+    replay_barrier $SINGLEMDS
+    touch $MOUNT2/$tfile
+    touch $MOUNT1/$tfile-2
+    umount $MOUNT2
+    facet_failover $SINGLEMDS
+    umount -f $MOUNT1
+    zconf_mount `hostname` $MOUNT1 || error "mount1 fais"
+    zconf_mount `hostname` $MOUNT2 || error "mount2 fais"
+    checkstat $MOUNT1/$tfile-2 && return 1
+    checkstat $MOUNT2/$tfile && return 2
+    return 0
+}
+run_test 0b "lost client during waiting for next transno"
+
 test_1() {
     touch $MOUNT1/a
     replay_barrier $SINGLEMDS
@@ -215,7 +270,7 @@ test_12() {
     do_facet $SINGLEMDS lctl set_param fail_loc=0x80000302
     facet_failover $SINGLEMDS
     do_facet $SINGLEMDS lctl set_param fail_loc=0
-    df $MOUNT || return 1
+    clients_up || return 1
 
     ls $DIR/$tfile
     kill -USR1 $MULTIPID || return 3
@@ -236,11 +291,11 @@ test_13() {
     kill -USR1 $MULTIPID || return 3
     wait $MULTIPID || return 4
 
-    # drop close 
+    # drop close
     do_facet $SINGLEMDS lctl set_param fail_loc=0x80000115
     facet_failover $SINGLEMDS
     do_facet $SINGLEMDS lctl set_param fail_loc=0
-    df $MOUNT || return 1
+    clients_up || return 1
 
     ls $DIR/$tfile
     $CHECKSTAT -t file $DIR/$tfile || return 2
@@ -250,34 +305,50 @@ test_13() {
 }
 run_test 13 "close resend timeout"
 
-test_14() {
+# test 14a removed after 18143 because it shouldn't fail anymore and do the same
+# as test_15a
+
+test_14b() {
+    wait_mds_ost_sync
+    wait_destroy_complete
+    BEFOREUSED=`df -P $DIR | tail -1 | awk '{ print $3 }'`
+    mkdir -p $MOUNT1/$tdir
+    $SETSTRIPE -i 0 $MOUNT1/$tdir
     replay_barrier $SINGLEMDS
-    createmany -o $MOUNT1/$tfile- 25
-    createmany -o $MOUNT2/$tfile-2- 1
-    createmany -o $MOUNT1/$tfile-3- 25
+    createmany -o $MOUNT1/$tdir/$tfile- 5
+
+    $SETSTRIPE -i 0 $MOUNT2/f14b-3
+    echo "data" > $MOUNT2/f14b-3
+    createmany -o $MOUNT1/$tdir/$tfile-3- 5
     umount $MOUNT2
 
-    facet_failover $SINGLEMDS
-    # expect failover to fail due to missing client 2
-    df $MOUNT && return 1
-    sleep 1
+    fail $SINGLEMDS
+    wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
 
-    # first 25 files should have been replayed 
-    unlinkmany $MOUNT1/$tfile- 25 || return 2
+    # first 25 files should have been replayed
+    unlinkmany $MOUNT1/$tdir/$tfile- 5 || return 2
+    unlinkmany $MOUNT1/$tdir/$tfile-3- 5 || return 3
+
+    zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
 
-    zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail" 
+    wait_mds_ost_sync || return 4
+    wait_destroy_complete || return 5
+
+    AFTERUSED=`df -P $DIR | tail -1 | awk '{ print $3 }'`
+    log "before $BEFOREUSED, after $AFTERUSED"
+    [ $AFTERUSED -ne $BEFOREUSED ] && \
+        error "after $AFTERUSED > before $BEFOREUSED" && return 4
     return 0
 }
-run_test 14 "timeouts waiting for lost client during replay"
+run_test 14b "delete ost orphans if gap occured in objids due to VBR"
 
-test_15a() {	# was test_15
+test_15a() { # was test_15
     replay_barrier $SINGLEMDS
     createmany -o $MOUNT1/$tfile- 25
     createmany -o $MOUNT2/$tfile-2- 1
     umount $MOUNT2
 
-    facet_failover $SINGLEMDS
-    df $MOUNT || return 1
+    fail $SINGLEMDS
 
     unlinkmany $MOUNT1/$tfile- 25 || return 2
     [ -e $MOUNT1/$tfile-2-0 ] && error "$tfile-2-0 exists"
@@ -290,14 +361,12 @@ run_test 15a "timeout waiting for lost client during replay, 1 client completes"
 test_15c() {
     replay_barrier $SINGLEMDS
     for ((i = 0; i < 2000; i++)); do
-	echo "data" > "$MOUNT2/${tfile}-$i" || error "create ${tfile}-$i failed"
+        echo "data" > "$MOUNT2/${tfile}-$i" || error "create ${tfile}-$i failed"
     done
-    
     umount $MOUNT2
-    facet_failover $SINGLEMDS
 
-    df $MOUNT || return 1
-    
+    fail $SINGLEMDS
+
     zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
     return 0
 }
@@ -311,8 +380,7 @@ test_16() {
 
     facet_failover $SINGLEMDS
     sleep $TIMEOUT
-    facet_failover $SINGLEMDS
-    df $MOUNT || return 1
+    fail $SINGLEMDS
 
     unlinkmany $MOUNT1/$tfile- 25 || return 2
 
@@ -334,8 +402,7 @@ test_17() {
 
     facet_failover ost1
     sleep $TIMEOUT
-    facet_failover ost1
-    df $MOUNT || return 1
+    fail ost1
 
     unlinkmany $MOUNT1/$tfile- 25 || return 2
 
@@ -388,8 +455,7 @@ test_20() { #16389
     touch $MOUNT1/a
     touch $MOUNT2/b
     umount $MOUNT2
-    facet_failover $SINGLEMDS
-    df $MOUNT1 || return 1
+    fail $SINGLEMDS
     rm $MOUNT1/a
     zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
     TIER1=$((`date +%s` - BEFORE))
@@ -398,8 +464,7 @@ test_20() { #16389
     touch $MOUNT1/a
     touch $MOUNT2/b
     umount $MOUNT2
-    facet_failover $SINGLEMDS
-    df $MOUNT1 || return 1
+    fail $SINGLEMDS
     rm $MOUNT1/a
     zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
     TIER2=$((`date +%s` - BEFORE))
@@ -492,8 +557,16 @@ test_21b() {
     COS=0
     do_facet mds$num lctl set_param mdt.*.commit_on_sharing=$COS
 
-    test_21b_sub mds$num && error "Not all renames are replayed. COS=$COS" 
-
+    # there is still a window when transactions may be written to disk before
+    # the mds device is set R/O. To avoid such a rare test failure, the check
+    # is repeated several times.
+    local n_attempts=1
+    while true; do
+    test_21b_sub mds$num || break;
+    let n_attempts=n_attempts+1
+    [ $n_attempts -gt 3 ] &&
+        error "The test cannot check whether COS works or not: all renames are replied w/o COS"
+    done
     restore_lustre_params < $param_file
     rm -f $param_file
     return 0
@@ -502,9 +575,9 @@ run_test 21b "commit on sharing, two clients"
 
 # end commit on sharing tests 
 
-equals_msg `basename $0`: test complete, cleaning up
+complete $(basename $0) $SECONDS
 SLEEP=$((`date +%s` - $NOW))
 [ $SLEEP -lt $TIMEOUT ] && sleep $SLEEP
 [ "$MOUNTED2" = yes ] && zconf_umount $HOSTNAME $MOUNT2 || true
 check_and_cleanup_lustre
-[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true
+exit_status