X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=9877521213a88465c53d568d02a2449d30129b61;hp=42bb2fe12aefebc3b019fbac77d8d0bba39c9f29;hb=98ceaf854bb4738305769c5cd1df556ee99aa859;hpb=3cce65712d94cffe8f1626545845b95b88aef672 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 42bb2fe..9877521 100644 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -53,16 +53,16 @@ fi if [[ $(uname -m) = aarch64 ]]; then # bug number: LU-11596 ALWAYS_EXCEPT+=" $GRANT_CHECK_LIST" - # bug number: LU-11671 LU-11667 LU-11729 LU-4398 - ALWAYS_EXCEPT+=" 45 317 810 817" + # bug number: LU-11671 LU-11667 LU-4398 + ALWAYS_EXCEPT+=" 45 317 817" fi # 5 12 (min)" [ "$SLOW" = "no" ] && EXCEPT_SLOW="27m 64b 68 71 115 300o" if [ "$mds1_FSTYPE" = "zfs" ]; then - # bug number for skipped test: LU-1957 - ALWAYS_EXCEPT="$ALWAYS_EXCEPT 180" + # bug number for skipped test: + ALWAYS_EXCEPT="$ALWAYS_EXCEPT " # 13 (min)" [ "$SLOW" = "no" ] && EXCEPT_SLOW="$EXCEPT_SLOW 51b" fi @@ -6807,9 +6807,9 @@ test_60a() { local pass=true #get fid and record list - fid_list=($(awk '/9_sub.*record/ { print $NF }' /$TMP/$tfile | + fid_list=($(awk '/9_sub.*record/ { print $NF }' $TMP/$tfile | tail -n 4)) - rec_list=($(awk '/9_sub.*record/ { print $((NF-3)) }' /$TMP/$tfile | + rec_list=($(awk '/9_sub.*record/ { print $((NF-3)) }' $TMP/$tfile | tail -n 4)) #remount mgs as ldiskfs or zfs type stop mgs || error "stop mgs failed" @@ -6928,13 +6928,16 @@ run_test 60e "no space while new llog is being created" test_60g() { local pid + local i test_mkdir -c $MDSCOUNT $DIR/$tdir - $LFS setdirstripe -D -i -1 -c $MDSCOUNT $DIR/$tdir ( local index=0 while true; do + $LFS setdirstripe -i $(($index % $MDSCOUNT)) \ + -c $MDSCOUNT $DIR/$tdir/subdir$index \ + 2>/dev/null mkdir $DIR/$tdir/subdir$index 2>/dev/null rmdir $DIR/$tdir/subdir$index 2>/dev/null index=$((index + 1)) @@ -6943,16 +6946,34 @@ test_60g() { pid=$! - for i in $(seq 100); do + for i in {0..100}; do # define OBD_FAIL_OSD_TXN_START 0x19a - do_facet mds1 lctl set_param fail_loc=0x8000019a + local index=$((i % MDSCOUNT + 1)) + + do_facet mds$index $LCTL set_param fail_loc=0x8000019a \ + > /dev/null usleep 100 done kill -9 $pid + for i in $(seq $MDSCOUNT); do + do_facet mds$i $LCTL set_param fail_loc=0 > /dev/null + done + mkdir $DIR/$tdir/new || error "mkdir failed" rmdir $DIR/$tdir/new || error "rmdir failed" + + do_facet mds1 $LCTL lfsck_start -M $(facet_svc mds1) -A -C \ + -t namespace + for i in $(seq $MDSCOUNT); do + wait_update_facet mds$i "$LCTL get_param -n \ + mdd.$(facet_svc mds$i).lfsck_namespace | + awk '/^status/ { print \\\$2 }'" "completed" + done + + ls -R $DIR/$tdir || error "ls failed" + rm -rf $DIR/$tdir || error "rmdir failed" } run_test 60g "transaction abort won't cause MDT hung" @@ -8444,12 +8465,14 @@ test_101c() { cancel_lru_locks osc $LCTL set_param osc.*.rpc_stats 0 $READS -f $DIR/$tfile -s$FILE_LENGTH -b$rsize -n$nreads -t 180 + $LCTL get_param osc.*.rpc_stats for osc_rpc_stats in $($LCTL get_param -N osc.*.rpc_stats); do local stats=$($LCTL get_param -n $osc_rpc_stats) local lines=$(echo "$stats" | awk 'END {print NR;}') local size if [ $lines -le 20 ]; then + echo "continue debug" continue fi for size in 1 2 4 8; do @@ -13748,6 +13771,49 @@ test_160i() { } run_test 160i "changelog user register/unregister race" +test_160j() { + remote_mds_nodsh && skip "remote MDS with nodsh" + [[ $MDS1_VERSION -lt $(version_code 2.12.56) ]] && + skip "Need MDS version at least 2.12.56" + + mount_client $MOUNT2 || error "mount_client on $MOUNT2 failed" + + changelog_register || error "first changelog_register failed" + + # generate some changelog + test_mkdir -c $MDSCOUNT $DIR/$tdir || error "mkdir $tdir failed" + createmany -m $DIR/$tdir/${tfile}bis $((MDSCOUNT * 2)) || + error "create $DIR/$tdir/${tfile}bis failed" + + # open the changelog device + exec 3>/dev/changelog-$FSNAME-MDT0000 + exec 4/dev/null || error "read changelog failed" + + # clear changelog + local cl_user="${CL_USERS[$SINGLEMDS]%% *}" + changelog_users $SINGLEMDS | grep -q $cl_user || + error "User $cl_user not found in changelog_users" + + printf 'clear:'$cl_user':0' >&3 + + # close + exec 3>&- + exec 4<&- + + # cleanup + changelog_deregister || error "changelog_deregister failed" + + umount $MOUNT2 + mount_client $MOUNT || error "mount_client on $MOUNT failed" +} +run_test 160j "client can be umounted while its chanangelog is being used" + test_161a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" @@ -16818,7 +16884,7 @@ test_243() } run_test 243 "various group lock tests" -test_244() +test_244a() { test_mkdir $DIR/$tdir dd if=/dev/zero of=$DIR/$tdir/$tfile bs=1M count=35 @@ -16826,7 +16892,26 @@ test_244() error "sendfile+grouplock failed" rm -rf $DIR/$tdir } -run_test 244 "sendfile with group lock tests" +run_test 244a "sendfile with group lock tests" + +test_244b() +{ + [ $PARALLEL == "yes" ] && skip "skip parallel run" && return + + local threads=50 + local size=$((1024*1024)) + + test_mkdir $DIR/$tdir + for i in $(seq 1 $threads); do + local file=$DIR/$tdir/file_$((i / 10)) + $MULTIOP $file OG1234w$size_$((i % 3))w$size_$((i % 4))g1234c & + local pids[$i]=$! + done + for i in $(seq 1 $threads); do + wait ${pids[$i]} + done +} +run_test 244b "multi-threaded write with group lock" test_245() { local flagname="multi_mod_rpcs" @@ -18235,8 +18320,8 @@ test_271f() { local mdtidx=$($LFS getstripe --mdt-index $DIR/$tdir) cancel_lru_locks mdc - dd if=/dev/urandom of=$tmp bs=200000 count=1 - dd if=$tmp of=$dom bs=200000 count=1 + dd if=/dev/urandom of=$tmp bs=265000 count=1 + dd if=$tmp of=$dom bs=265000 count=1 cancel_lru_locks mdc cat /etc/hosts >> $tmp lctl set_param -n mdc.*.stats=clear @@ -18263,6 +18348,7 @@ test_271f() { local ra=$(get_mdc_stats $mdtidx req_active) local rw=$(get_mdc_stats $mdtidx req_waittime) + [ -z $num ] && num=0 [ $num -eq 1 ] || error "expect 1 READ RPC, $num occured" [ $ra == $rw ] || error "$((ra - rw)) resend occured" echo "... DONE" @@ -18342,12 +18428,12 @@ test_272b() { $LFS migrate -c2 $dom || error "failed to migrate to the new composite layout" - [ $($LFS getstripe -L $dom) == 'mdt' ] && + [ $($LFS getstripe -L $dom) != 'mdt' ] || error "MDT stripe was not removed" cancel_lru_locks mdc local new_md5=$(md5sum $dom) - [ "$old_md5" != "$new_md5" ] && + [ "$old_md5" == "$new_md5" ] || error "$old_md5 != $new_md5" # Skip free space checks with ZFS @@ -18387,7 +18473,7 @@ test_272c() { cancel_lru_locks mdc local new_md5=$(md5sum $dom) - [ "$old_md5" != "$new_md5" ] && + [ "$old_md5" == "$new_md5" ] || error "$old_md5 != $new_md5" # Skip free space checks with ZFS @@ -18401,6 +18487,108 @@ test_272c() { } run_test 272c "DoM migration: DOM file to the OST-striped file (composite)" +test_272d() { + [ $MDS1_VERSION -lt $(version_code 2.12.55) ] && + skip "Need MDS version at least 2.12.55" + + local dom=$DIR/$tdir/$tfile + mkdir -p $DIR/$tdir + $LFS setstripe -E 1M -L mdt -E -1 -c1 $dom + + local mdtidx=$($LFS getstripe -m $dom) + local mdtname=MDT$(printf %04x $mdtidx) + local facet=mds$((mdtidx + 1)) + + dd if=/dev/urandom of=$dom bs=2M count=1 oflag=direct || + error "failed to write data into $dom" + local old_md5=$(md5sum $dom) + cancel_lru_locks mdc + local mdtfree1=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + + $LFS mirror extend -N -E 2M -c1 -E -1 -c2 $dom || + error "failed mirroring to the new composite layout" + $LFS mirror resync $dom || + error "failed mirror resync" + $LFS mirror split --mirror-id 1 -d $dom || + error "failed mirror split" + + [ $($LFS getstripe -L $dom) != 'mdt' ] || + error "MDT stripe was not removed" + + cancel_lru_locks mdc + local new_md5=$(md5sum $dom) + [ "$old_md5" == "$new_md5" ] || + error "$old_md5 != $new_md5" + + # Skip free space checks with ZFS + if [ "$(facet_fstype $facet)" != "zfs" ]; then + local mdtfree2=$(do_facet $facet \ + lctl get_param -n osd*.*$mdtname.kbytesfree) + [ $mdtfree2 -gt $mdtfree1 ] || + error "MDS space is not freed after DOM mirror deletion" + fi + return 0 +} +run_test 272d "DoM mirroring: OST-striped mirror to DOM file" + +test_272e() { + [ $MDS1_VERSION -lt $(version_code 2.12.55) ] && + skip "Need MDS version at least 2.12.55" + + local dom=$DIR/$tdir/$tfile + mkdir -p $DIR/$tdir + $LFS setstripe -c 2 $dom + + dd if=/dev/urandom of=$dom bs=512K count=1 oflag=direct || + error "failed to write data into $dom" + local old_md5=$(md5sum $dom) + cancel_lru_locks mdc + + $LFS mirror extend -N -E 1M -L mdt -E eof -c2 $dom || + error "failed mirroring to the DOM layout" + $LFS mirror resync $dom || + error "failed mirror resync" + $LFS mirror split --mirror-id 1 -d $dom || + error "failed mirror split" + + [ $($LFS getstripe -L $dom) != 'mdt' ] || + error "MDT stripe was not removed" + + cancel_lru_locks mdc + local new_md5=$(md5sum $dom) + [ "$old_md5" == "$new_md5" ] || + error "$old_md5 != $new_md5" + + return 0 +} +run_test 272e "DoM mirroring: DOM mirror to the OST-striped file" + +test_272f() { + [ $MDS1_VERSION -lt $(version_code 2.12.55) ] && + skip "Need MDS version at least 2.12.55" + + local dom=$DIR/$tdir/$tfile + mkdir -p $DIR/$tdir + $LFS setstripe -c 2 $dom + + dd if=/dev/urandom of=$dom bs=512K count=1 oflag=direct || + error "failed to write data into $dom" + local old_md5=$(md5sum $dom) + cancel_lru_locks mdc + + $LFS migrate -E 1M -L mdt -E eof -c2 -v $dom || + error "failed migrating to the DOM file" + + cancel_lru_locks mdc + local new_md5=$(md5sum $dom) + [ "$old_md5" != "$new_md5" ] && + error "$old_md5 != $new_md5" + + return 0 +} +run_test 272f "DoM migration: OST-striped file to DOM file" + test_273a() { [ $MDS1_VERSION -lt $(version_code 2.11.50) ] && skip "Need MDS version at least 2.11.50" @@ -21558,20 +21746,29 @@ test_809() { run_test 809 "Verify no SOM xattr store for DoM-only files" test_810() { - local ORIG - local CSUM - - # t10 seem to dislike partial pages - lctl set_param osc.*.checksum_type=adler - lctl set_param fail_loc=0x411 - dd if=/dev/urandom of=$DIR/$tfile bs=10240 count=2 - ORIG=$(md5sum $DIR/$tfile) - lctl set_param ldlm.namespaces.*osc*.lru_size=clear - CSUM=$(md5sum $DIR/$tfile) - set_checksum_type adler - if [ "$ORIG" != "$CSUM" ]; then - error "$ORIG != $CSUM" - fi + [ $PARALLEL == "yes" ] && skip "skip parallel run" + $GSS && skip_env "could not run with gss" + + set_checksums 1 + stack_trap "set_checksums $ORIG_CSUM" EXIT + stack_trap "set_checksum_type $ORIG_CSUM_TYPE" EXIT + + local csum + local before + local after + for csum in $CKSUM_TYPES; do + #define OBD_FAIL_OSC_NO_GRANT 0x411 + $LCTL set_param osc.*.checksum_type=$csum fail_loc=0x411 + for i in "10240 0" "10000 0" "4000 1" "500 1"; do + eval set -- $i + dd if=/dev/urandom of=$DIR/$tfile bs=$1 count=2 seek=$2 + before=$(md5sum $DIR/$tfile) + $LCTL set_param ldlm.namespaces.*osc*.lru_size=clear + after=$(md5sum $DIR/$tfile) + [ "$before" == "$after" ] || + error "$csum: $before != $after bs=$1 seek=$2" + done + done } run_test 810 "partial page writes on ZFS (LU-11663)" @@ -21850,6 +22047,19 @@ test_817() { } run_test 817 "nfsd won't cache write lock for exec file" +test_818() { + mkdir $DIR/$tdir + $LFS setstripe -c1 -i0 $DIR/$tfile + $LFS setstripe -c1 -i1 $DIR/$tfile + stop $SINGLEMDS + #define OBD_FAIL_OSP_CANT_PROCESS_LLOG 0x2105 + do_facet $SINGLEMDS lctl set_param fail_loc=0x80002105 + start $SINGLEMDS $(mdsdevname ${SINGLEMDS//mds/}) $MDS_MOUNT_OPTS || + error "start $SINGLEMDS failed" + rm -rf $DIR/$tdir +} +run_test 818 "unlink with failed llog" + # # tests that do cleanup/setup should be run at the end #