X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Freplay-dual.sh;h=8f0cb527ac0a2c9374a1efa51007865edfe47d99;hb=dfd8433e6f6376bf5f7a50020bb484adfe48ce2c;hp=2f1646722b1b3a1ffcf97c4a3333e73491660e49;hpb=0fe560ea5dd4e582f2393f54de106e34a8b2c6ae;p=fs%2Flustre-release.git

diff --git a/lustre/tests/replay-dual.sh b/lustre/tests/replay-dual.sh
index 2f16467..8f0cb52 100755
--- a/lustre/tests/replay-dual.sh
+++ b/lustre/tests/replay-dual.sh
@@ -2,27 +2,45 @@
 
 set -e
 
-# bug number:  13129 13129 6088 10124 
-ALWAYS_EXCEPT="2     3     8    15c   $REPLAY_DUAL_EXCEPT"
+# bug number:  10124
+ALWAYS_EXCEPT="15c   $REPLAY_DUAL_EXCEPT"
 
+LFS=${LFS:-lfs}
+SETSTRIPE=${SETSTRIPE:-"$LFS setstripe"}
+GETSTRIPE=${GETSTRIPE:-"$LFS getstripe"}
 SAVE_PWD=$PWD
 PTLDEBUG=${PTLDEBUG:--1}
-LUSTRE=${LUSTRE:-`dirname $0`/..}
+LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
 SETUP=${SETUP:-""}
 CLEANUP=${CLEANUP:-""}
 MOUNT_2=${MOUNT_2:-"yes"}
 . $LUSTRE/tests/test-framework.sh
 
-init_test_env $@
+if [ "$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
 
-#
-[ "$SLOW" = "no" ] && EXCEPT_SLOW="1 2 3 4 5 14"
+[ "$SLOW" = "no" ] && EXCEPT_SLOW="21b"
 
 build_test_filter
 
-cleanup_and_setup_lustre
+check_and_setup_lustre
+MOUNTED=$(mounted_lustre_filesystems)
+if ! $(echo $MOUNTED | grep -w -q $MOUNT2); then
+    zconf_mount $HOSTNAME $MOUNT2
+    MOUNTED2=yes
+fi
+
+assert_DIR
 rm -rf $DIR/[df][0-9]*
 
 [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
@@ -200,7 +218,7 @@ 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
@@ -221,11 +239,11 @@ 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
@@ -235,7 +253,7 @@ test_13() {
 }
 run_test 13 "close resend timeout"
 
-test_14() {
+test_14a() {
     replay_barrier $SINGLEMDS
     createmany -o $MOUNT1/$tfile- 25
     createmany -o $MOUNT2/$tfile-2- 1
@@ -244,7 +262,7 @@ test_14() {
 
     facet_failover $SINGLEMDS
     # expect failover to fail due to missing client 2
-    df $MOUNT && return 1
+    client_evicted || return 1
     sleep 1
 
     # first 25 files should have been replayed 
@@ -253,101 +271,67 @@ test_14() {
     zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail" 
     return 0
 }
-run_test 14 "timeouts waiting for lost client during replay"
+run_test 14a "timeouts waiting for lost client during replay"
 
-test_15() {
+test_14b() {
+    wait_mds_ost_sync
+    wait_destroy_complete
+    BEFOREUSED=`df -P $DIR | tail -1 | awk '{ print $3 }'`
+    mkdir -p $MOUNT1/$tdir
+    $SETSTRIPE -o 0 $MOUNT1/$tdir
     replay_barrier $SINGLEMDS
-    createmany -o $MOUNT1/$tfile- 25
-    createmany -o $MOUNT2/$tfile-2- 1
+    createmany -o $MOUNT1/$tdir/$tfile- 5
+
+    $SETSTRIPE -o 0 $MOUNT2/f14b-3
+    echo "data" > $MOUNT2/f14b-3
+    createmany -o $MOUNT1/$tdir/$tfile-3- 5
     umount $MOUNT2
 
-    facet_failover $SINGLEMDS
-    df $MOUNT || return 1
+    fail $SINGLEMDS
+    wait_recovery_complete $SINGLEMDS || error "MDS recovery not done"
 
-    unlinkmany $MOUNT1/$tfile- 25 || return 2
-    [ -e $MOUNT1/$tfile-2-0 ] && error "$tfile-2-0 exists"
+    # 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"
-    return 0
-}
-run_test 15 "timeout waiting for lost client during replay, 1 client completes"
-
-test_15a() {
-    local ost_last_id=""
-    local osc_last_id=""
-    
-    replay_barrier $SINGLEMDS
-    echo "data" > "$MOUNT2/${tfile}-m2"
 
-    umount $MOUNT2
-    facet_failover $SINGLEMDS
-    df $MOUNT || return 1
-    
-    ost_last_id=`lctl get_param -n obdfilter.*.last_id`
-    mds_last_id=`lctl get_param -n osc.*mds*.last_id`
-    
-    echo "Ids after MDS<->OST synchonizing"
-    echo "--------------------------------"
-    echo "MDS last_id:"
-    echo $mds_last_id
-    echo "OST last_id:"
-    echo $ost_last_id
-
-    local i=0
-    echo $ost_last_id | while read id; do
-	ost_ids[$i]=$id
-	((i++))
-    done
-    
-    i=0
-    echo $mds_last_id | while read id; do
-	mds_ids[$i]=$id
-	((i++))
-    done
-    
-    local arr_len=${#mds_ids[*]}
-    for ((i=0;i<$arr_len;i++)); do
-	    mds_id=${mds_ids[i]}
-	    ost_id=${ost_ids[i]}
-	    
-	    test $mds_id -ge $ost_id || {
-		echo "MDS last id ($mds_id) is smaller than OST one ($ost_id)"
-		return 2
-	    }
-    done
+    wait_mds_ost_sync || return 4
+    wait_destroy_complete || return 5
 
-    zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
+    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
 }
-#CROW run_test 15a "OST clear orphans - synchronize ids on MDS and OST"
+run_test 14b "delete ost orphans if gap occured in objids due to VBR"
 
-test_15b() {
+test_15a() { # was test_15
     replay_barrier $SINGLEMDS
-    echo "data" > "$MOUNT2/${tfile}-m2"
+    createmany -o $MOUNT1/$tfile- 25
+    createmany -o $MOUNT2/$tfile-2- 1
     umount $MOUNT2
 
-    do_facet ost1 "lctl set_param fail_loc=0x80000802"
-    facet_failover $SINGLEMDS
+    fail $SINGLEMDS
+
+    unlinkmany $MOUNT1/$tfile- 25 || return 2
+    [ -e $MOUNT1/$tfile-2-0 ] && error "$tfile-2-0 exists"
 
-    df $MOUNT || return 1
-    do_facet ost1 "lctl set_param fail_loc=0"
-    
     zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
     return 0
 }
-#CROW run_test 15b "multiple delayed OST clear orphans"
+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
 }
@@ -361,8 +345,7 @@ test_16() {
 
     facet_failover $SINGLEMDS
     sleep $TIMEOUT
-    facet_failover $SINGLEMDS
-    df $MOUNT || return 1
+    fail $SINGLEMDS
 
     unlinkmany $MOUNT1/$tfile- 25 || return 2
 
@@ -373,6 +356,8 @@ test_16() {
 run_test 16 "fail MDS during recovery (3571)"
 
 test_17() {
+    remote_ost_nodsh && skip "remote OST with nodsh" && return 0
+
     createmany -o $MOUNT1/$tfile- 25
     createmany -o $MOUNT2/$tfile-2- 1
 
@@ -382,8 +367,7 @@ test_17() {
 
     facet_failover ost1
     sleep $TIMEOUT
-    facet_failover ost1
-    df $MOUNT || return 1
+    fail ost1
 
     unlinkmany $MOUNT1/$tfile- 25 || return 2
 
@@ -430,9 +414,135 @@ 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 $SINGLEMDS
+    touch $MOUNT1/a
+    touch $MOUNT2/b
+    umount $MOUNT2
+    fail $SINGLEMDS
+    rm $MOUNT1/a
+    zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
+    TIER1=$((`date +%s` - BEFORE))
+    BEFORE=`date +%s`
+    replay_barrier $SINGLEMDS
+    touch $MOUNT1/a
+    touch $MOUNT2/b
+    umount $MOUNT2
+    fail $SINGLEMDS
+    rm $MOUNT1/a
+    zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
+    TIER2=$((`date +%s` - BEFORE))
+    [ $TIER2 -ge $((TIER1 * 2)) ] && \
+        error "recovery time is growing $TIER2 > $TIER1"
+    return 0
+}
+run_test 20 "recovery time is not increasing"
+
+# commit on sharing tests
+test_21a() {
+    local param_file=$TMP/$tfile-params
+
+    save_lustre_params $(facet_active_host $SINGLEMDS) "mdt.*.commit_on_sharing" > $param_file
+    do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=1
+    touch  $MOUNT1/$tfile-1
+    mv  $MOUNT2/$tfile-1 $MOUNT2/$tfile-2
+    mv  $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
+    replay_barrier_nosync $SINGLEMDS
+    umount $MOUNT2
+
+    facet_failover $SINGLEMDS
+
+    # all renames are replayed
+    unlink  $MOUNT1/$tfile-3 || return 2
+
+    zconf_mount `hostname` $MOUNT2 || error "mount $MOUNT2 fail"
+
+    do_facet $SINGLEMDS lctl set_param mdt.*.commit_on_sharing=0
+    rm -rf $MOUNT1/$tfile-*
+    restore_lustre_params < $param_file
+    rm -f $param_file
+    return 0
+}
+run_test 21a "commit on sharing"
+
+test_21b_sub () {
+    local mds=$1 
+    do_node $CLIENT1 rm -f $MOUNT1/$tfile-*
+
+    do_facet $mds sync
+    do_node $CLIENT1 touch  $MOUNT1/$tfile-1
+    do_node $CLIENT2 mv  $MOUNT1/$tfile-1 $MOUNT1/$tfile-2
+    do_node $CLIENT1 mv  $MOUNT1/$tfile-2 $MOUNT1/$tfile-3
+
+    replay_barrier_nosync $mds
+    shutdown_client $CLIENT2 $MOUNT1
+
+    facet_failover $mds
+
+    # were renames replayed?
+    local rc=0
+    echo UNLINK $MOUNT1/$tfile-3 
+    do_node $CLIENT1 unlink  $MOUNT1/$tfile-3 || \
+        { echo "unlink $tfile-3 fail!" && rc=1; }
+
+    boot_node $CLIENT2
+    zconf_mount_clients $CLIENT2 $MOUNT1 || error "mount $CLIENT2 $MOUNT1 fail" 
+
+    return $rc
+}
+
+test_21b() {
+    [ -z "$CLIENTS" ] && skip "Need two or more clients." && return
+    [ $CLIENTCOUNT -lt 2 ] && \
+        { skip "Need two or more clients, have $CLIENTCOUNT" && return; }
+
+    if [ "$FAILURE_MODE" = "HARD" ] &&  mixed_mdt_devs; then
+        skip "Several mdt services on one mds node are used with FAILURE_MODE=$FAILURE_MODE. "
+        return 0
+    fi
+
+
+    zconf_umount_clients $CLIENTS $MOUNT2
+    zconf_mount_clients $CLIENTS $MOUNT1
+
+    local param_file=$TMP/$tfile-params
+
+    local num=$(get_mds_dir $MOUNT1)
+
+    save_lustre_params $(facet_active_host mds$num) "mdt.*.commit_on_sharing" > $param_file
+
+    # COS enabled
+    local COS=1
+    do_facet mds$num lctl set_param mdt.*.commit_on_sharing=$COS
+
+    test_21b_sub mds$num || error "Not all renames are replayed. COS=$COS"
+
+    # COS disabled (should fail)
+    COS=0
+    do_facet mds$num lctl set_param mdt.*.commit_on_sharing=$COS
+
+    # there is still a window when transactions may be written to disk before
+    # the mds device is set R/O. To avoid such a rare test failure, the check
+    # is repeated several times.
+    local n_attempts=1
+    while true; do
+    test_21b_sub mds$num || break;
+    let n_attempts=n_attempts+1
+    [ $n_attemtps -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
+}
+run_test 21b "commit on sharing, two clients"
+
+# end commit on sharing tests 
+
 equals_msg `basename $0`: test complete, cleaning up
 SLEEP=$((`date +%s` - $NOW))
 [ $SLEEP -lt $TIMEOUT ] && sleep $SLEEP
+[ "$MOUNTED2" = yes ] && zconf_umount $HOSTNAME $MOUNT2 || true
 check_and_cleanup_lustre
-[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG || true
-
+[ -f "$TESTSUITELOG" ] && cat $TESTSUITELOG && grep -q FAIL $TESTSUITELOG && exit 1 || true