X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fsanity.sh;h=0f32b1a9dbb085bf88bb18bc6027920f3800b4e0;hp=39241192a03036b5546ea50f1a7e8ea36059bc39;hb=ea58c4cfb0fc255befbbb7754bd4ed71704a2a2c;hpb=d4a54de84c0554a21b215621b4df5fdd93e38f74 diff --git a/lustre/tests/sanity.sh b/lustre/tests/sanity.sh index 3924119..0f32b1a 100644 --- a/lustre/tests/sanity.sh +++ b/lustre/tests/sanity.sh @@ -45,24 +45,30 @@ ALWAYS_EXCEPT+=" 42a 42b 42c " ALWAYS_EXCEPT+=" 407 312 " if $SHARED_KEY; then - # bug number: LU-9795 LU-9795 LU-9795 LU-9795 - ALWAYS_EXCEPT+=" 17n 60a 133g 300f " + # bug number: LU-9795 LU-9795 LU-9795 LU-9795 LU-12781 + ALWAYS_EXCEPT+=" 17n 60a 133g 300f 272a" +fi + +selinux_status=$(getenforce) +if [ "$selinux_status" != "Disabled" ]; then + # bug number: LU-12895 LU-12469 LU-12469 + ALWAYS_EXCEPT+=" 185a 230b 230d" fi # skip the grant tests for ARM until they are fixed 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="$EXCEPT_SLOW 51b" fi @@ -2458,10 +2464,6 @@ test_27D() { local ost_list=$(seq $first_ost $ost_step $last_ost) local ost_range="$first_ost $last_ost $ost_step" - if ! combined_mgs_mds ; then - mount_mgs_client - fi - test_mkdir $DIR/$tdir pool_add $POOL || error "pool_add failed" pool_add_targets $POOL $ost_range || error "pool_add_targets failed" @@ -2472,19 +2474,15 @@ test_27D() { [ $MDS1_VERSION -lt $(version_code 2.9.55) ] || [ $CLIENT_VERSION -lt $(version_code 2.9.55) ] && skip27D+=" -s 30,31" - [ $(lustre_version_code $SINGLEMDS) -lt $(version_code $SEL_VER) ] && - skip27D+="-s 32" [[ ! $($LCTL get_param mdc.*.import) =~ connect_flags.*overstriping || $OSTCOUNT -ge $(($LOV_MAX_STRIPE_COUNT / 2)) ]] && skip27D+=" -s 32,33" + [[ $MDS_VERSION -lt $(version_code $SEL_VER) ]] && + skip27D+=" -s 34" llapi_layout_test -d$DIR/$tdir -p$POOL -o$OSTCOUNT $skip27D || error "llapi_layout_test failed" destroy_test_pools || error "destroy test pools failed" - - if ! combined_mgs_mds ; then - umount_mgs_client - fi } run_test 27D "validate llapi_layout API" @@ -2620,6 +2618,8 @@ run_test 27H "Set specific OSTs stripe" test_27I() { [ $PARALLEL == "yes" ] && skip "skip parallel run" [[ $OSTCOUNT -lt 2 ]] && skip_env "needs >= 2 OSTs" + [[ $MDS1_VERSION -gt $(version_code 2.12.52) ]] || + skip "Need MDS version newer than 2.12.52" local pool=$TESTNAME local ostrange="1 1 1" @@ -2815,11 +2815,6 @@ test_27L() { local POOL=${POOL:-$TESTNAME} - if ! combined_mgs_mds ; then - mount_mgs_client - trap umount_mgs_client EXIT - fi - pool_add $POOL || error "pool_add failed" lfs pool_list $MOUNT | grep -Fx "${FSNAME}.${POOL}" || @@ -2828,6 +2823,152 @@ test_27L() { } run_test 27L "lfs pool_list gives correct pool name" +test_27M() { + [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.12.57) ]] && + skip "Need MDS version >= than 2.12.57" + remote_mds_nodsh && skip "remote MDS with nodsh" + [[ $OSTCOUNT -lt 2 ]] && skip_env "need > 1 OST" + + test_mkdir $DIR/$tdir + + # Set default striping on directory + $LFS setstripe -C 4 $DIR/$tdir + + echo 1 > $DIR/$tdir/${tfile}.1 + local count=$($LFS getstripe -c $DIR/$tdir/${tfile}.1) + local setcount=4 + [ $count -eq $setcount ] || + error "(1) stripe count $count, should be $setcount" + + # Capture existing append_stripe_count setting for restore + local orig_count=$(do_facet mds1 $LCTL get_param -n mdd.$FSNAME-MDT0000.append_stripe_count) + local mdts=$(comma_list $(mdts_nodes)) + stack_trap "do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=$orig_count" EXIT + + local appendcount=$orig_count + echo 1 >> $DIR/$tdir/${tfile}.2_append + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.2_append) + [ $count -eq $appendcount ] || + error "(2)stripe count $count, should be $appendcount for append" + + # Disable O_APPEND striping, verify it works + do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=0 + + # Should now get the default striping, which is 4 + setcount=4 + echo 1 >> $DIR/$tdir/${tfile}.3_append + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.3_append) + [ $count -eq $setcount ] || + error "(3) stripe count $count, should be $setcount" + + # Try changing the stripe count for append files + do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=2 + + # Append striping is now 2 (directory default is still 4) + appendcount=2 + echo 1 >> $DIR/$tdir/${tfile}.4_append + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.4_append) + [ $count -eq $appendcount ] || + error "(4) stripe count $count, should be $appendcount for append" + + # Test append stripe count of -1 + do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=-1 + appendcount=$OSTCOUNT + echo 1 >> $DIR/$tdir/${tfile}.5 + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.5) + [ $count -eq $appendcount ] || + error "(5) stripe count $count, should be $appendcount for append" + + # Set append striping back to default of 1 + do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=1 + + # Try a new default striping, PFL + DOM + $LFS setstripe -L mdt -E 1M -E -1 -c 2 $DIR/$tdir + + # Create normal DOM file, DOM returns stripe count == 0 + setcount=0 + touch $DIR/$tdir/${tfile}.6 + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.6) + [ $count -eq $setcount ] || + error "(6) stripe count $count, should be $setcount" + + # Show + appendcount=1 + echo 1 >> $DIR/$tdir/${tfile}.7_append + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.7_append) + [ $count -eq $appendcount ] || + error "(7) stripe count $count, should be $appendcount for append" + + # Clean up DOM layout + $LFS setstripe -d $DIR/$tdir + + # Now test that append striping works when layout is from root + $LFS setstripe -c 2 $MOUNT + # Make a special directory for this + mkdir $DIR/${tdir}/${tdir}.2 + stack_trap "$LFS setstripe -d $MOUNT" EXIT + + # Verify for normal file + setcount=2 + echo 1 > $DIR/${tdir}/${tdir}.2/${tfile}.8 + count=$($LFS getstripe -c $DIR/$tdir/${tdir}.2/${tfile}.8) + [ $count -eq $setcount ] || + error "(8) stripe count $count, should be $setcount" + + appendcount=1 + echo 1 >> $DIR/${tdir}/${tdir}.2/${tfile}.9_append + count=$($LFS getstripe -c $DIR/${tdir}/${tdir}.2/${tfile}.9_append) + [ $count -eq $appendcount ] || + error "(9) stripe count $count, should be $appendcount for append" + + # Now test O_APPEND striping with pools + do_nodes $mdts $LCTL set_param mdd.*.append_pool="$TESTNAME" + stack_trap "do_nodes $mdts $LCTL set_param mdd.*.append_pool='none'" EXIT + + # Create the pool + pool_add $TESTNAME || error "pool creation failed" + pool_add_targets $TESTNAME 0 1 || error "Pool add targets failed" + + echo 1 >> $DIR/$tdir/${tfile}.10_append + + pool=$($LFS getstripe -p $DIR/$tdir/${tfile}.10_append) + [ "$pool" = "$TESTNAME" ] || error "(10) incorrect pool: $pool" + + # Check that count is still correct + appendcount=1 + echo 1 >> $DIR/$tdir/${tfile}.11_append + count=$($LFS getstripe -c $DIR/$tdir/${tfile}.11_append) + [ $count -eq $appendcount ] || + error "(11) stripe count $count, should be $appendcount for append" + + # Disable O_APPEND stripe count, verify pool works separately + do_nodes $mdts $LCTL set_param mdd.*.append_stripe_count=0 + + echo 1 >> $DIR/$tdir/${tfile}.12_append + + pool=$($LFS getstripe -p $DIR/$tdir/${tfile}.12_append) + [ "$pool" = "$TESTNAME" ] || error "(12) incorrect pool: $pool" + + # Remove pool setting, verify it's not applied + do_nodes $mdts $LCTL set_param mdd.*.append_pool='none' + + echo 1 >> $DIR/$tdir/${tfile}.13_append + + pool=$($LFS getstripe -p $DIR/$tdir/${tfile}.13_append) + [ "$pool" = "" ] || error "(13) pool found: $pool" +} +run_test 27M "test O_APPEND striping" + +test_27N() { + combined_mgs_mds && skip "needs separate MGS/MDT" + + pool_add $TESTNAME || error "pool_add failed" + do_facet mgs "$LCTL pool_list $FSNAME" | + grep -Fx "${FSNAME}.${TESTNAME}" || + error "lctl pool_list on MGS failed" +} +run_test 27N "lctl pool_list on separate MGS gives correct pool name" + # createtest also checks that device nodes are created and # then visible correctly (#2091) test_28() { # bug 2091 @@ -4133,7 +4274,7 @@ test_39j() { error "mtime is lost on close: $mtime2, " \ "should be $mtime1" - cancel_lru_locks $OSC + cancel_lru_locks if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi done lctl set_param fail_loc=0 @@ -4162,7 +4303,7 @@ test_39k() { [ "$mtime2" = $TEST_39_MTIME ] || \ error "mtime is lost on close: $mtime2, should be $TEST_39_MTIME" - cancel_lru_locks osc + cancel_lru_locks if [ $i = 0 ] ; then echo "repeat after cancel_lru_locks"; fi done } @@ -5360,19 +5501,27 @@ test_56c() { skip_env "OST $ost_name is in $old_status status" do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=1 + [[ $OST1_VERSION -ge $(version_code 2.12.55) ]] && do_facet ost1 \ + $LCTL set_param -n obdfilter.$ost_name.no_precreate=1 sleep_maxage local new_status=$(ost_dev_status $ost_idx) - [[ "$new_status" = "D" ]] || - error "OST $ost_name is in status of '$new_status', not 'D'" + [[ "$new_status" =~ "D" ]] || + error "$ost_name status is '$new_status', missing 'D'" + if [[ $OST1_VERSION -ge $(version_code 2.12.55) ]]; then + [[ "$new_status" =~ "N" ]] || + error "$ost_name status is '$new_status', missing 'N'" + fi do_facet ost1 $LCTL set_param -n obdfilter.$ost_name.degraded=0 + [[ $OST1_VERSION -ge $(version_code 2.12.55) ]] && do_facet ost1 \ + $LCTL set_param -n obdfilter.$ost_name.no_precreate=0 sleep_maxage new_status=$(ost_dev_status $ost_idx) - [[ -z "$new_status" ]] || - error "OST $ost_name is in status of '$new_status', not ''" + [[ ! "$new_status" =~ "D" && ! "$new_status" =~ "N" ]] || + error "$ost_name status is '$new_status', has 'D' and/or 'N'" } run_test 56c "check 'lfs df' showing device status" @@ -5649,39 +5798,73 @@ test_56r() { setup_56 $dir $NUMFILES $NUMDIRS local expected=12 - local cmd="$LFS find -size 0 -type f $dir" + local cmd="$LFS find -size 0 -type f -lazy $dir" local nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + cmd="$LFS find -size 0 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + expected=0 + cmd="$LFS find ! -size 0 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find ! -size 0 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + echo "test" > $dir/$tfile echo "test2" > $dir/$tfile.2 && sync expected=1 + cmd="$LFS find -size 5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find -size 5 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + expected=1 + cmd="$LFS find -size +5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find -size +5 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + expected=2 + cmd="$LFS find -size +0 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find -size +0 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + expected=2 + cmd="$LFS find ! -size -5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find ! -size -5 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || error "'$cmd' wrong: found $nums, expected $expected" + expected=12 + cmd="$LFS find -size -5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" cmd="$LFS find -size -5 -type f $dir" nums=$($cmd | wc -l) [ $nums -eq $expected ] || @@ -5689,6 +5872,157 @@ test_56r() { } run_test 56r "check lfs find -size works" +test_56ra() { + [[ $MDS1_VERSION -gt $(version_code 2.12.58) ]] || + skip "MDS < 2.12.58 doesn't return LSOM data" + local dir=$DIR/$tdir + + [[ $OSC == "mdc" ]] && skip "DoM files" && return + + setup_56 $dir $NUMFILES $NUMDIRS "-c 1" + + cancel_lru_locks $OSC + + local rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + local expected=12 + local cmd="$LFS find -size 0 -type f -lazy $dir" + local nums=$($cmd | wc -l) + + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + + local rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find -size 0 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + $LCTL get_param osc.*.stats + [ $rpcs_after -eq $((rpcs_before + 12)) ] || + error "'$cmd' should send 12 glimpse RPCs to OST" + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=0 + cmd="$LFS find ! -size 0 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + $LCTL get_param mdc.*.stats + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find ! -size 0 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 12)) ] || + error "'$cmd' should send 12 glimpse RPCs to OST" + + echo "test" > $dir/$tfile + echo "test2" > $dir/$tfile.2 && sync + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=1 + cmd="$LFS find -size 5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find -size 5 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 14)) ] || + error "'$cmd' should send 14 glimpse RPCs to OST" + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=1 + cmd="$LFS find -size +5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find -size +5 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 14)) ] || + error "'$cmd' should send 14 glimpse RPCs to OST" + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=2 + cmd="$LFS find -size +0 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find -size +0 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 14)) ] || + error "'$cmd' should send 14 glimpse RPCs to OST" + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=2 + cmd="$LFS find ! -size -5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find ! -size -5 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 14)) ] || + error "'$cmd' should send 14 glimpse RPCs to OST" + + cancel_lru_locks $OSC + rpcs_before=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + expected=12 + cmd="$LFS find -size -5 -type f -lazy $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + [ $rpcs_before -eq $rpcs_after ] || + error "'$cmd' should not send glimpse RPCs to OST" + cmd="$LFS find -size -5 -type f $dir" + nums=$($cmd | wc -l) + [ $nums -eq $expected ] || + error "'$cmd' wrong: found $nums, expected $expected" + rpcs_after=$(calc_stats $OSC.*$OSC*.stats ldlm_glimpse_enqueue) + echo "Before: $rpcs_before After: $rpcs_after $NUMFILES" + [ $rpcs_after -eq $((rpcs_before + 14)) ] || + error "'$cmd' should send 14 glimpse RPCs to OST" +} +run_test 56ra "check lfs find -size -lazy works for data on OSTs" + test_56s() { # LU-611 #LU-9369 [[ $OSTCOUNT -lt 2 ]] && skip_env "need at least 2 OSTs" @@ -5911,7 +6245,7 @@ test_56w() { done # $LFS_MIGRATE will fail if hard link migration is unsupported - if [[ $(lustre_version_code mds1) -gt $(version_code 2.5.55) ]]; then + if [[ $MDS1_VERSION -gt $(version_code 2.5.55) ]]; then createmany -l$dir/dir1/file1 $dir/dir1/link 200 || error "creating links to $dir/dir1/file1 failed" fi @@ -6249,6 +6583,7 @@ check_migrate_links() { local file1="$dir/file1" local begin="$2" local count="$3" + local runas="$4" local total_count=$(($begin + $count - 1)) local symlink_count=10 local uniq_count=10 @@ -6293,7 +6628,7 @@ check_migrate_links() { fi echo -n "migrating files..." - local migrate_out=$($LFS_MIGRATE -y -S '1m' $dir) + local migrate_out=$($runas $LFS_MIGRATE -y -S '1m' $dir) local rc=$? [ $rc -eq 0 ] || error "migrate failed rc = $rc" echo "done" @@ -6348,6 +6683,9 @@ test_56xb() { echo "testing rsync mode when all links do not fit within xattrs" LFS_MIGRATE_RSYNC_MODE=true check_migrate_links "$dir" 101 100 + chown -R $RUNAS_ID $dir + echo "testing non-root lfs migrate mode when not all links are in xattr" + LFS_MIGRATE_RSYNC_MODE=false check_migrate_links "$dir" 101 100 "$RUNAS" # clean up rm -rf $dir @@ -6807,9 +7145,9 @@ test_60a() { local pass=true #get fid and record list - fid_list=($(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" @@ -7525,6 +7863,29 @@ test_65n() { local dir5_layout=$(get_layout_param $dir5) [[ "$dir4_layout" = "$dir5_layout" ]] || error "$dir5 should inherit the default layout from $dir4" + + # though subdir under ROOT doesn't inherit default layout, but + # its sub dir/file should be created with default layout. + [[ $MDSCOUNT -ge 2 ]] || skip_env "needs >= 2 MDTs" + [[ $MDS1_VERSION -ge $(version_code 2.12.59) ]] || + skip "Need MDS version at least 2.12.59" + + local default_lmv_count=$($LFS getdirstripe -D -c $MOUNT) + local default_lmv_index=$($LFS getdirstripe -D -i $MOUNT) + local default_lmv_hash=$($LFS getdirstripe -D -H $MOUNT) + + if [ $default_lmv_hash == "none" ]; then + stack_trap "$LFS setdirstripe -D -d $MOUNT" EXIT + else + stack_trap "$LFS setdirstripe -D -i $default_lmv_index \ + -c $default_lmv_count -H $default_lmv_hash $MOUNT" EXIT + fi + + $LFS setdirstripe -D -c 2 $MOUNT || + error "setdirstripe -D -c 2 failed" + mkdir $MOUNT/$tdir-6 || error "mkdir $tdir-6 failed" + local lmv_count=$($LFS getdirstripe -c $MOUNT/$tdir-6) + [ $lmv_count -eq 2 ] || error "$tdir-6 stripe count $lmv_count" } run_test 65n "don't inherit default layout from root for new subdirectories" @@ -8012,7 +8373,7 @@ test_77k() { # LU-10906 wait_update $HOSTNAME "$get_checksum" $i #remount echo "remount client, checksum should be $i" - remount_client $MOUNT || "failed to remount client" + remount_client $MOUNT || error "failed to remount client" checksum=$(eval $get_checksum) [ $checksum -eq $i ] || error "checksum($checksum) != $i" done @@ -8023,15 +8384,15 @@ test_77k() { # LU-10906 for opt in "checksum" "nochecksum"; do #remount with mount option echo "remount client with option $opt, checksum should be $i" - umount_client $MOUNT || "failed to umount client" + umount_client $MOUNT || error "failed to umount client" mount_client $MOUNT "$MOUNT_OPTS,$opt" || - "failed to mount client with option '$opt'" + error "failed to mount client with option '$opt'" checksum=$(eval $get_checksum) [ $checksum -eq $i ] || error "checksum($checksum) != $i" i=$((i - 1)) done - remount_client $MOUNT || "failed to remount client" + remount_client $MOUNT || error "failed to remount client" } run_test 77k "enable/disable checksum correctly" @@ -8328,7 +8689,6 @@ cleanup_101a() { test_101a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" - [ $MDSCOUNT -ge 2 ] && skip_env "needs < 2 MDTs" #LU-4322 local s local discard @@ -8353,9 +8713,12 @@ test_101a() { done cleanup_101a - if [[ $(($discard * 10)) -gt $nreads ]]; then - $LCTL get_param osc.*-osc*.rpc_stats - $LCTL get_param llite.*.read_ahead_stats + $LCTL get_param osc.*-osc*.rpc_stats + $LCTL get_param llite.*.read_ahead_stats + + # Discard is generally zero, but sometimes a few random reads line up + # and trigger larger readahead, which is wasted & leads to discards. + if [[ $(($discard)) -gt $nreads ]]; then error "too many ($discard) discarded pages" fi rm -f $DIR/$tfile || true @@ -10998,41 +11361,52 @@ run_test 126 "check that the fsgid provided by the client is taken into account" test_127a() { # bug 15521 [ $PARALLEL == "yes" ] && skip "skip parallel run" + local name count samp unit min max sum sumsq $LFS setstripe -i 0 -c 1 $DIR/$tfile || error "setstripe failed" + echo "stats before reset" + $LCTL get_param osc.*.stats $LCTL set_param osc.*.stats=0 - FSIZE=$((2048 * 1024)) - dd if=/dev/zero of=$DIR/$tfile bs=$FSIZE count=1 + local fsize=$((2048 * 1024)) + + dd if=/dev/zero of=$DIR/$tfile bs=$fsize count=1 cancel_lru_locks osc - dd if=$DIR/$tfile of=/dev/null bs=$FSIZE - - $LCTL get_param osc.*0000-osc-*.stats | grep samples > $DIR/${tfile}.tmp - while read NAME COUNT SAMP UNIT MIN MAX SUM SUMSQ; do - echo "got $COUNT $NAME" - [ ! $MIN ] && error "Missing min value for $NAME proc entry" - eval $NAME=$COUNT || error "Wrong proc format" - - case $NAME in - read_bytes|write_bytes) - [ $MIN -lt 4096 ] && error "min is too small: $MIN" - [ $MIN -gt $FSIZE ] && error "min is too big: $MIN" - [ $MAX -lt 4096 ] && error "max is too small: $MAX" - [ $MAX -gt $FSIZE ] && error "max is too big: $MAX" - [ $SUM -ne $FSIZE ] && error "sum is wrong: $SUM" - [ $SUMSQ -lt $(((FSIZE /4096) * (4096 * 4096))) ] && - error "sumsquare is too small: $SUMSQ" - [ $SUMSQ -gt $((FSIZE * FSIZE)) ] && - error "sumsquare is too big: $SUMSQ" - ;; - *) ;; - esac - done < $DIR/${tfile}.tmp - - #check that we actually got some stats - [ "$read_bytes" ] || error "Missing read_bytes stats" - [ "$write_bytes" ] || error "Missing write_bytes stats" - [ "$read_bytes" != 0 ] || error "no read done" - [ "$write_bytes" != 0 ] || error "no write done" + dd if=$DIR/$tfile of=/dev/null bs=$fsize + + $LCTL get_param osc.*0000-osc-*.stats | grep samples > $DIR/$tfile.tmp + stack_trap "rm -f $TMP/$tfile.tmp" + while read name count samp unit min max sum sumsq; do + echo "got name=$name count=$count unit=$unit min=$min max=$max" + [ ! $min ] && error "Missing min value for $name proc entry" + eval $name=$count || error "Wrong proc format" + + case $name in + read_bytes|write_bytes) + [[ "$unit" =~ "bytes" ]] || + error "unit is not 'bytes': $unit" + (( $min >= 4096 )) || error "min is too small: $min" + (( $min <= $fsize )) || error "min is too big: $min" + (( $max >= 4096 )) || error "max is too small: $max" + (( $max <= $fsize )) || error "max is too big: $max" + (( $sum == $fsize )) || error "sum is wrong: $sum" + (( $sumsq >= ($fsize / 4096) * (4096 * 4096) )) || + error "sumsquare is too small: $sumsq" + (( $sumsq <= $fsize * $fsize )) || + error "sumsquare is too big: $sumsq" + ;; + ost_read|ost_write) + [[ "$unit" =~ "usec" ]] || + error "unit is not 'usec': $unit" + ;; + *) ;; + esac + done < $DIR/$tfile.tmp + + #check that we actually got some stats + [ "$read_bytes" ] || error "Missing read_bytes stats" + [ "$write_bytes" ] || error "Missing write_bytes stats" + [ "$read_bytes" != 0 ] || error "no read done" + [ "$write_bytes" != 0 ] || error "no write done" } run_test 127a "verify the client stats are sane" @@ -11040,6 +11414,8 @@ test_127b() { # bug LU-333 [ $PARALLEL == "yes" ] && skip "skip parallel run" local name count samp unit min max sum sumsq + echo "stats before reset" + $LCTL get_param llite.*.stats $LCTL set_param llite.*.stats=0 # perform 2 reads and writes so MAX is different from SUM. @@ -11050,30 +11426,28 @@ test_127b() { # bug LU-333 dd if=$DIR/$tfile of=/dev/null bs=$PAGE_SIZE count=1 $LCTL get_param llite.*.stats | grep samples > $TMP/$tfile.tmp + stack_trap "rm -f $TMP/$tfile.tmp" while read name count samp unit min max sum sumsq; do - echo "got $count $name" + echo "got name=$name count=$count unit=$unit min=$min max=$max" eval $name=$count || error "Wrong proc format" case $name in - read_bytes) - [ $count -ne 2 ] && error "count is not 2: $count" - [ $min -ne $PAGE_SIZE ] && + read_bytes|write_bytes) + [[ "$unit" =~ "bytes" ]] || + error "unit is not 'bytes': $unit" + (( $count == 2 )) || error "count is not 2: $count" + (( $min == $PAGE_SIZE )) || error "min is not $PAGE_SIZE: $min" - [ $max -ne $PAGE_SIZE ] && - error "max is incorrect: $max" - [ $sum -ne $((PAGE_SIZE * 2)) ] && - error "sum is wrong: $sum" + (( $max == $PAGE_SIZE )) || + error "max is not $PAGE_SIZE: $max" + (( $sum == $PAGE_SIZE * 2 )) || + error "sum is not $((PAGE_SIZE * 2)): $sum" ;; - write_bytes) - [ $count -ne 2 ] && error "count is not 2: $count" - [ $min -ne $PAGE_SIZE ] && - error "min is not $PAGE_SIZE: $min" - [ $max -ne $PAGE_SIZE ] && - error "max is incorrect: $max" - [ $sum -ne $((PAGE_SIZE * 2)) ] && - error "sum is wrong: $sum" + read|write) + [[ "$unit" =~ "usec" ]] || + error "unit is not 'usec': $unit" ;; - *) ;; + *) ;; esac done < $TMP/$tfile.tmp @@ -11082,8 +11456,6 @@ test_127b() { # bug LU-333 [ "$write_bytes" ] || error "Missing write_bytes stats" [ "$read_bytes" != 0 ] || error "no read done" [ "$write_bytes" != 0 ] || error "no write done" - - rm -f $TMP/${tfile}.tmp } run_test 127b "verify the llite client stats are sane" @@ -13771,6 +14143,77 @@ 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_160k() { + [ $PARALLEL == "yes" ] && skip "skip parallel run" + remote_mds_nodsh && skip "remote MDS with nodsh" + + mkdir -p $DIR/$tdir/1/1 + + changelog_register || error "changelog_register failed" + local cl_user="${CL_USERS[$SINGLEMDS]%% *}" + + changelog_users $SINGLEMDS | grep -q $cl_user || + error "User '$cl_user' not found in changelog_users" +#define OBD_FAIL_MDS_CHANGELOG_REORDER 0x15d + do_facet mds1 $LCTL set_param fail_loc=0x8000015d fail_val=3 + rmdir $DIR/$tdir/1/1 & sleep 1 + mkdir $DIR/$tdir/2 + touch $DIR/$tdir/2/2 + rm -rf $DIR/$tdir/2 + + wait + sleep 4 + + changelog_dump | grep rmdir || error "rmdir not recorded" + + rm -rf $DIR/$tdir + changelog_deregister +} +run_test 160k "Verify that changelog records are not lost" + test_161a() { [ $PARALLEL == "yes" ] && skip "skip parallel run" @@ -14723,10 +15166,6 @@ test_200() { local subdir=$test_path/subdir local rc=0 - if ! combined_mgs_mds ; then - mount_mgs_client - fi - while : ; do # former test_200a test_200b pool_add $POOL || { rc=$? ; break; } @@ -14760,9 +15199,6 @@ test_200() { destroy_test_pools - if ! combined_mgs_mds ; then - umount_mgs_client - fi return $rc } run_test 200 "OST pools" @@ -15444,10 +15880,6 @@ test_220() { #LU-325 $LFS df -i - if ! combined_mgs_mds ; then - mount_mgs_client - fi - do_facet ost$((OSTIDX + 1)) lctl set_param fail_val=-1 #define OBD_FAIL_OST_ENOINO 0x229 do_facet ost$((OSTIDX + 1)) lctl set_param fail_loc=0x229 @@ -15484,10 +15916,6 @@ test_220() { #LU-325 error "$LCTL pool_destroy $FSNAME.$TESTNAME failed" echo "unlink $MDSOBJS files @$next_id..." unlinkmany $DIR/$tdir/f $MDSOBJS || error "unlinkmany failed" - - if ! combined_mgs_mds ; then - umount_mgs_client - fi } run_test 220 "preallocated MDS objects still used if ENOSPC from OST" @@ -16841,7 +17269,7 @@ 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 @@ -16849,7 +17277,26 @@ 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" @@ -17149,9 +17596,6 @@ test_253() { wait_delete_completed mkdir $DIR/$tdir - if ! combined_mgs_mds ; then - mount_mgs_client - fi pool_add $TESTNAME || error "Pool creation failed" pool_add_targets $TESTNAME 0 || error "Pool add targets failed" @@ -17197,10 +17641,6 @@ test_253() { if (( oa_status != 0 )); then error "Object allocation still disable after rm" fi - - if ! combined_mgs_mds ; then - umount_mgs_client - fi } run_test 253 "Check object allocation limit" @@ -17506,7 +17946,7 @@ run_test 255b "check 'lfs ladvise -a dontneed'" test_255c() { [ $OST1_VERSION -lt $(version_code 2.10.50) ] && - skip "lustre < 2.10.53 does not support lockahead" + skip "lustre < 2.10.50 does not support lockahead" local count local new_count @@ -18258,8 +18698,8 @@ test_271f() { local mdtidx=$($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 >> $tmp lctl set_param -n mdc.*.stats=clear @@ -18286,6 +18726,7 @@ test_271f() { local ra=$(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" @@ -18365,12 +18806,12 @@ 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 != $new_md5" # Skip free space checks with ZFS @@ -18410,7 +18851,7 @@ test_272c() { cancel_lru_locks mdc local new_md5=$(md5sum $dom) - [ "$old_md5" != "$new_md5" ] && + [ "$old_md5" == "$new_md5" ] || error "$old_md5 != $new_md5" # Skip free space checks with ZFS @@ -18424,6 +18865,108 @@ 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" @@ -18540,6 +19083,7 @@ test_278() { do_facet mds2 $LCTL set_param fail_loc=0 start mds2 $(mdsdevname 2) $MDS_MOUNT_OPTS + wait_recovery_complete mds2 } run_test 278 "Race starting MDS between MDTs stop/start" @@ -20019,9 +20563,6 @@ test_406() { local def_stripe_size=$($LFS getstripe -S $MOUNT) local test_pool=$TESTNAME - if ! combined_mgs_mds ; then - mount_mgs_client - fi pool_add $test_pool || error "pool_add failed" pool_add_targets $test_pool 0 $(($OSTCOUNT - 1)) 1 || error "pool_add_targets failed" @@ -20074,10 +20615,6 @@ test_406() { local f=$DIR/$tdir/$tfile pool_remove_all_targets $test_pool $f pool_remove $test_pool $f - - if ! combined_mgs_mds ; then - umount_mgs_client - fi } run_test 406 "DNE support fs default striping" @@ -20228,174 +20765,247 @@ test_412() { } run_test 412 "mkdir on specific MDTs" -test_413a() { - [ $MDSCOUNT -lt 2 ] && - skip "We need at least 2 MDTs for this test" - - if [ $(lustre_version_code mds1) -lt $(version_code 2.10.55) ]; then - skip "Need server version at least 2.10.55" - fi - - mkdir $DIR/$tdir || error "mkdir failed" - - # find MDT that is the most full - local max=$($LFS df | grep MDT | - awk 'BEGIN { a=0 } - { sub("%", "", $5) - if (0+$5 >= a) - { - a = $5 - b = $6 - } - } - END { split(b, c, ":") - sub("]", "", c[2]) - print c[2] - }') - - for i in $(seq $((MDSCOUNT - 1))); do - $LFS mkdir -c $i $DIR/$tdir/d$i || - error "mkdir d$i failed" - $LFS getdirstripe $DIR/$tdir/d$i - local stripe_index=$($LFS getdirstripe -i $DIR/$tdir/d$i) - [ $stripe_index -ne $max ] || - error "don't expect $max" - done -} -run_test 413a "mkdir on less full MDTs" - -test_413b() { - [ $MDSCOUNT -lt 2 ] && - skip "We need at least 2 MDTs for this test" - - [ $MDS1_VERSION -lt $(version_code 2.12.52) ] && - skip "Need server version at least 2.12.52" - - mkdir $DIR/$tdir || error "mkdir failed" - $LFS setdirstripe -D -i -1 -H space $DIR/$tdir || - error "setdirstripe failed" +test_qos_mkdir() { + local mkdir_cmd=$1 + local stripe_count=$2 + local mdts=$(comma_list $(mdts_nodes)) - local qos_prio_free - local qos_threshold_rr + local testdir + local lmv_qos_prio_free + local lmv_qos_threshold_rr + local lmv_qos_maxage + local lod_qos_prio_free + local lod_qos_threshold_rr + local lod_qos_maxage local count + local i - qos_prio_free=$($LCTL get_param -n lmv.*.qos_prio_free | head -n1) - qos_prio_free=${qos_prio_free%%%} - qos_threshold_rr=$($LCTL get_param -n lmv.*.qos_threshold_rr | head -n1) - qos_threshold_rr=${qos_threshold_rr%%%} - qos_maxage=$($LCTL get_param -n lmv.*.qos_maxage) - - stack_trap "$LCTL set_param lmv.*.qos_prio_free=$qos_prio_free" EXIT - stack_trap "$LCTL set_param lmv.*.qos_threshold_rr=$qos_threshold_rr" \ + lmv_qos_prio_free=$($LCTL get_param -n lmv.*.qos_prio_free | head -n1) + lmv_qos_prio_free=${lmv_qos_prio_free%%%} + lmv_qos_threshold_rr=$($LCTL get_param -n lmv.*.qos_threshold_rr | + head -n1) + lmv_qos_threshold_rr=${lmv_qos_threshold_rr%%%} + lmv_qos_maxage=$($LCTL get_param -n lmv.*.qos_maxage) + stack_trap "$LCTL set_param \ + lmv.*.qos_prio_free=$lmv_qos_prio_free > /dev/null" EXIT + stack_trap "$LCTL set_param \ + lmv.*.qos_threshold_rr=$lmv_qos_threshold_rr > /dev/null" EXIT + stack_trap "$LCTL set_param \ + lmv.*.qos_maxage=$lmv_qos_maxage > /dev/null" EXIT + + lod_qos_prio_free=$(do_facet mds1 $LCTL get_param -n \ + lod.lustre-MDT0000-mdtlov.mdt_qos_prio_free | head -n1) + lod_qos_prio_free=${lod_qos_prio_free%%%} + lod_qos_threshold_rr=$(do_facet mds1 $LCTL get_param -n \ + lod.lustre-MDT0000-mdtlov.mdt_qos_threshold_rr | head -n1) + lod_qos_threshold_rr=${lod_qos_threshold_rr%%%} + lod_qos_maxage=$(do_facet mds1 $LCTL get_param -n \ + lod.lustre-MDT0000-mdtlov.qos_maxage | awk '{ print $1 }') + stack_trap "do_nodes $mdts $LCTL set_param \ + lod.*.mdt_qos_prio_free=$lod_qos_prio_free > /dev/null" EXIT + stack_trap "do_nodes $mdts $LCTL set_param \ + lod.*.mdt_qos_threshold_rr=$lod_qos_threshold_rr > /dev/null" \ EXIT - stack_trap "$LCTL set_param lmv.*.qos_maxage=$qos_maxage" EXIT + stack_trap "do_nodes $mdts $LCTL set_param \ + lod.*.mdt_qos_maxage=$lod_qos_maxage > /dev/null" EXIT + + echo + echo "Mkdir (stripe_count $stripe_count) roundrobin:" + + $LCTL set_param lmv.*.qos_threshold_rr=100 > /dev/null + do_nodes $mdts $LCTL set_param lod.*.mdt_qos_threshold_rr=100 > /dev/null - echo "mkdir with roundrobin" + testdir=$DIR/$tdir-s$stripe_count/rr - $LCTL set_param lmv.*.qos_threshold_rr=100 for i in $(seq $((100 * MDSCOUNT))); do - mkdir $DIR/$tdir/subdir$i || error "mkdir subdir$i failed" + eval $mkdir_cmd $testdir/subdir$i || + error "$mkdir_cmd subdir$i failed" done + for i in $(seq $MDSCOUNT); do - count=$($LFS getdirstripe -i $DIR/$tdir/* | grep ^$((i - 1))$ | - wc -w) + count=$($LFS getdirstripe -i $testdir/* | + grep ^$((i - 1))$ | wc -l) echo "$count directories created on MDT$((i - 1))" [ $count -eq 100 ] || error "subdirs are not evenly distributed" + + if [ $stripe_count -gt 1 ]; then + count=$($LFS getdirstripe $testdir/* | + grep -P "^\s+$((i - 1))\t" | wc -l) + echo "$count stripes created on MDT$((i - 1))" + # deviation should < 5% of average + [ $count -lt $((95 * stripe_count)) ] || + [ $count -gt $((105 * stripe_count)) ] && + error "stripes are not evenly distributed" + fi done - rm -rf $DIR/$tdir/* + $LCTL set_param lmv.*.qos_threshold_rr=$lmv_qos_threshold_rr > /dev/null + do_nodes $mdts $LCTL set_param \ + lod.*.mdt_qos_threshold_rr=$lod_qos_threshold_rr > /dev/null - $LCTL set_param lmv.*.qos_threshold_rr=$qos_threshold_rr - # Shorten statfs result age, so that it can be updated in time - $LCTL set_param lmv.*.qos_maxage=1 - sleep_maxage + echo + echo "Check for uneven MDTs: " local ffree + local bavail local max local min local max_index local min_index + local tmp - ffree=($(lctl get_param -n mdc.*[mM][dD][cC]-[^M]*.filesfree | uniq)) - echo "MDT filesfree available: ${ffree[@]}" - max=${ffree[0]} - min=${ffree[0]} + ffree=($(lctl get_param -n mdc.*[mM][dD][cC]-[^M]*.filesfree)) + bavail=($(lctl get_param -n mdc.*[mM][dD][cC]-[^M]*.kbytesavail)) + bsize=$(lctl get_param -n mdc.*MDT0000*.blocksize) + + max=$(((${ffree[0]} >> 8) * (${bavail[0]} * bsize >> 16))) + min=$(((${ffree[0]} >> 8) * (${bavail[0]} * bsize >> 16))) max_index=0 min_index=0 - for ((i = 0; i < ${#ffree[@]}; i++)); do - if [[ ${ffree[i]} -gt $max ]]; then - max=${ffree[i]} + for ((i = 1; i < ${#ffree[@]}; i++)); do + tmp=$(((${ffree[i]} >> 8) * (${bavail[i]} * bsize >> 16))) + if [ $tmp -gt $max ]; then + max=$tmp max_index=$i fi - if [[ ${ffree[i]} -lt $min ]]; then - min=${ffree[i]} + if [ $tmp -lt $min ]; then + min=$tmp min_index=$i fi done - echo "Min free files: MDT$min_index: $min" - echo "Max free files: MDT$max_index: $max" - [ $min -eq 0 ] && skip "no free files in MDT$min_index" - [ $min -gt 10000000 ] && skip "too much free files in MDT$min_index" + [ ${ffree[min_index]} -eq 0 ] && + skip "no free files in MDT$min_index" + [ ${ffree[min_index]} -gt 100000000 ] && + skip "too much free files in MDT$min_index" # Check if we need to generate uneven MDTs - test_mkdir -i $min_index -c 1 -p $DIR/$tdir-MDT$min_index - local threshold=10 - local diff=$((max - min)) - local diff2=$((diff * 100 / min)) - - echo -n "Check for uneven MDTs: " - echo -n "diff=$diff files ($diff2%) must be > $threshold% ..." - - if [ $diff2 -gt $threshold ]; then - echo "ok" - echo "Don't need to fill MDT$min_index" - else - # generate uneven MDTs, create till 25% diff - echo "no" - diff2=$((threshold - diff2)) - diff=$((min * diff2 / 100)) + local threshold=50 + local diff=$(((max - min) * 100 / min)) + local value="$(generate_string 1024)" + + while [ $diff -lt $threshold ]; do + # generate uneven MDTs, create till $threshold% diff + echo -n "weight diff=$diff% must be > $threshold% ..." + count=$((${ffree[min_index]} / 10)) # 50 sec per 10000 files in vm - [ $diff -gt 40000 ] && [ "$SLOW" = "no" ] && - skip "$diff files to create" - echo "Fill $diff2% diff in MDT$min_index with $diff files" - local i - local value="$(generate_string 1024)" - for i in $(seq $diff); do + [ $count -gt 40000 ] && [ "$SLOW" = "no" ] && + skip "$count files to create" + echo "Fill MDT$min_index with $count files" + [ -d $DIR/$tdir-MDT$min_index ] || + $LFS mkdir -i $min_index $DIR/$tdir-MDT$min_index || + error "mkdir $tdir-MDT$min_index failed" + for i in $(seq $count); do $OPENFILE -f O_CREAT:O_LOV_DELAY_CREATE \ - $DIR/$tdir-MDT$min_index/f$i > /dev/null || - error "create f$i failed" + $DIR/$tdir-MDT$min_index/f$j_$i > /dev/null || + error "create f$j_$i failed" setfattr -n user.413b -v $value \ - $DIR/$tdir-MDT$min_index/f$i || - error "setfattr f$i failed" + $DIR/$tdir-MDT$min_index/f$j_$i || + error "setfattr f$j_$i failed" done - fi - min=$((100 *MDSCOUNT)) - max=0 + ffree=($(lctl get_param -n mdc.*[mM][dD][cC]-*.filesfree)) + bavail=($(lctl get_param -n mdc.*[mM][dD][cC]-*.kbytesavail)) + max=$(((${ffree[max_index]} >> 8) * \ + (${bavail[max_index]} * bsize >> 16))) + min=$(((${ffree[min_index]} >> 8) * \ + (${bavail[min_index]} * bsize >> 16))) + diff=$(((max - min) * 100 / min)) + done + + echo "MDT filesfree available: ${ffree[@]}" + echo "MDT blocks available: ${bavail[@]}" + echo "weight diff=$diff%" + + echo + echo "Mkdir (stripe_count $stripe_count) with balanced space usage:" + + $LCTL set_param lmv.*.qos_prio_free=100 > /dev/null + do_nodes $mdts $LCTL set_param lod.*.mdt_qos_prio_free=100 > /dev/null + # decrease statfs age, so that it can be updated in time + $LCTL set_param lmv.*.qos_maxage=1 > /dev/null + do_nodes $mdts $LCTL set_param lod.*.mdt_qos_maxage=1 > /dev/null + + sleep 1 + + testdir=$DIR/$tdir-s$stripe_count/qos - echo "mkdir with balanced space usage" - $LCTL set_param lmv.*.qos_prio_free=100 for i in $(seq $((100 * MDSCOUNT))); do - mkdir $DIR/$tdir/subdir$i || error "mkdir subdir$i failed" + eval $mkdir_cmd $testdir/subdir$i || + error "$mkdir_cmd subdir$i failed" done + for i in $(seq $MDSCOUNT); do - count=$($LFS getdirstripe -i $DIR/$tdir/* | grep ^$((i - 1))$ | - wc -w) + count=$($LFS getdirstripe -i $testdir/* | grep ^$((i - 1))$ | + wc -l) echo "$count directories created on MDT$((i - 1))" - [ $min -gt $count ] && min=$count - [ $max -lt $count ] && max=$count + + if [ $stripe_count -gt 1 ]; then + count=$($LFS getdirstripe $testdir/* | + grep -P "^\s+$((i - 1))\t" | wc -l) + echo "$count stripes created on MDT$((i - 1))" + fi done - [ $((max - min)) -gt $MDSCOUNT ] || + + max=$($LFS getdirstripe -i $testdir/* | grep ^$max_index$ | wc -l) + min=$($LFS getdirstripe -i $testdir/* | grep ^$min_index$ | wc -l) + + # D-value should > 10% of averge + [ $((max - min)) -lt 10 ] && error "subdirs shouldn't be evenly distributed" - which getfattr > /dev/null 2>&1 || skip_env "no getfattr command" + # ditto + if [ $stripe_count -gt 1 ]; then + max=$($LFS getdirstripe $testdir/* | + grep -P "^\s+$max_index\t" | wc -l) + min=$($LFS getdirstripe $testdir/* | + grep -P "^\s+$min_index\t" | wc -l) + [ $((max - min)) -le $((10 * stripe_count)) ] && + error "stripes shouldn't be evenly distributed"|| true + fi +} - $LFS setdirstripe -D -d $DIR/$tdir || error "setdirstripe -d failed" - getfattr -n trusted.dmv $DIR/$tdir && error "default dir layout exists" - true +test_413a() { + [ $MDSCOUNT -lt 2 ] && + skip "We need at least 2 MDTs for this test" + + [ $MDS1_VERSION -lt $(version_code 2.12.52) ] && + skip "Need server version at least 2.12.52" + + local stripe_count + + for stripe_count in $(seq 1 $((MDSCOUNT - 1))); do + mkdir $DIR/$tdir-s$stripe_count || error "mkdir failed" + mkdir $DIR/$tdir-s$stripe_count/rr || error "mkdir failed" + mkdir $DIR/$tdir-s$stripe_count/qos || error "mkdir failed" + test_qos_mkdir "$LFS mkdir -c $stripe_count" $stripe_count + done +} +run_test 413a "QoS mkdir with 'lfs mkdir -i -1'" + +test_413b() { + [ $MDSCOUNT -lt 2 ] && + skip "We need at least 2 MDTs for this test" + + [ $MDS1_VERSION -lt $(version_code 2.12.52) ] && + skip "Need server version at least 2.12.52" + + local stripe_count + + for stripe_count in $(seq 1 $((MDSCOUNT - 1))); do + mkdir $DIR/$tdir-s$stripe_count || error "mkdir failed" + mkdir $DIR/$tdir-s$stripe_count/rr || error "mkdir failed" + mkdir $DIR/$tdir-s$stripe_count/qos || error "mkdir failed" + $LFS setdirstripe -D -c $stripe_count \ + $DIR/$tdir-s$stripe_count/rr || + error "setdirstripe failed" + $LFS setdirstripe -D -c $stripe_count \ + $DIR/$tdir-s$stripe_count/qos || + error "setdirstripe failed" + test_qos_mkdir "mkdir" $stripe_count + done } -run_test 413b "mkdir with balanced space usage" +run_test 413b "QoS mkdir under dir whose default LMV starting MDT offset is -1" test_414() { #define OBD_FAIL_PTLRPC_BULK_ATTACH 0x521 @@ -20804,9 +21414,9 @@ test_421f() { cnt=$(ls -1 $DIR/$tdir | wc -l) [ $cnt == 1 ] || error "unexpected #files after (5): $cnt" - umount_client $MOUNT || "failed to umount client" + umount_client $MOUNT || error "failed to umount client" mount_client $MOUNT "$MOUNT_OPTS,user_fid2path" || - "failed to mount client'" + error "failed to mount client'" $RUNAS $LFS rmfid $DIR $FID || error "rmfid failed" # rmfid should succeed @@ -20820,9 +21430,9 @@ test_421f() { FID=$(lfs path2fid $DIR/$tdir/f) $RUNAS $LFS rmfid $DIR $FID && error "rmfid didn't fail" - umount_client $MOUNT || "failed to umount client" + umount_client $MOUNT || error "failed to umount client" mount_client $MOUNT "$MOUNT_OPTS" || - "failed to mount client'" + error "failed to mount client'" } run_test 421f "rmfid checks permissions" @@ -20859,6 +21469,50 @@ test_421g() { } run_test 421g "rmfid to return errors properly" +test_422() { + test_mkdir -i 0 -c 1 -p $DIR/$tdir/d1 + test_mkdir -i 0 -c 1 -p $DIR/$tdir/d2 + test_mkdir -i 0 -c 1 -p $DIR/$tdir/d3 + dd if=/dev/zero of=$DIR/$tdir/d1/file1 bs=1k count=1 + dd if=/dev/zero of=$DIR/$tdir/d2/file1 bs=1k count=1 + + local amc=$(at_max_get client) + local amo=$(at_max_get mds1) + local timeout=`lctl get_param -n timeout` + + at_max_set 0 client + at_max_set 0 mds1 + +#define OBD_FAIL_PTLRPC_PAUSE_REQ 0x50a + do_facet mds1 $LCTL set_param fail_loc=0x8000050a \ + fail_val=$(((2*timeout + 10)*1000)) + touch $DIR/$tdir/d3/file & + sleep 2 +#define OBD_FAIL_TGT_REPLY_DATA_RACE 0x722 + do_facet mds1 $LCTL set_param fail_loc=0x80000722 \ + fail_val=$((2*timeout + 5)) + mv $DIR/$tdir/d1/file1 $DIR/$tdir/d1/file2 & + local pid=$! + sleep 1 + kill -9 $pid + sleep $((2 * timeout)) + echo kill $pid + kill -9 $pid + lctl mark touch + touch $DIR/$tdir/d2/file3 + touch $DIR/$tdir/d2/file4 + touch $DIR/$tdir/d2/file5 + + wait + at_max_set $amc client + at_max_set $amo mds1 + + # LU-12838 - verify the ptlrpc thread watchdog is not always throttled + do_facet mds1 "dmesg | grep 'Dumping the stack trace for debugging'" || + error "Watchdog is always throttled" +} +run_test 422 "kill a process with RPC in progress" + prep_801() { [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] || [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] && @@ -21090,7 +21744,7 @@ cleanup_802a() { } test_802a() { - + [[ $mds1_FSTYPE = zfs ]] || skip "ZFS specific test" [[ $(lustre_version_code mds1) -lt $(version_code 2.9.55) ]] || [[ $OST1_VERSION -lt $(version_code 2.9.55) ]] && skip "Need server version at least 2.9.55" @@ -21306,7 +21960,7 @@ cleanup_805() { } test_805() { - local zfs_version=$(do_node $SINGLEMDS cat /sys/module/zfs/version) + local zfs_version=$(do_facet mds1 cat /sys/module/zfs/version) [ "$mds1_FSTYPE" != "zfs" ] && skip "ZFS specific test" [ $(version_code $zfs_version) -lt $(version_code 0.7.2) ] && skip "netfree not implemented before 0.7" @@ -21416,7 +22070,7 @@ test_806() { done check_lsom_size $DIR/$tfile $size - # multi-client wirtes + # multi-client writes num=$(get_node_count ${CLIENTS//,/ }) size=$(($num * $bs)) offset=0 @@ -21581,20 +22235,31 @@ test_809() { run_test 809 "Verify no SOM xattr store for DoM-only files" test_810() { - local ORIG - local CSUM - - # t10 seem to dislike partial pages - lctl set_param osc.*.checksum_type=adler - lctl set_param fail_loc=0x411 - dd if=/dev/urandom of=$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" + [[ $OST1_VERSION -gt $(version_code 2.12.58) ]] || + skip "OST < 2.12.58 doesn't align checksum" + + 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)" @@ -21873,6 +22538,39 @@ 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" + +test_819a() { + dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 + cancel_lru_locks osc + #define OBD_FAIL_OST_2BIG_NIOBUF 0x248 + do_facet $SINGLEMDS lctl set_param fail_loc=0x80000248 + dd if=$DIR/$tfile of=/dev/null bs=1M count=1 + rm -f $TDIR/$tfile +} +run_test 819a "too big niobuf in read" + +test_819b() { + #define OBD_FAIL_OST_2BIG_NIOBUF 0x248 + do_facet $SINGLEMDS lctl set_param fail_loc=0x80000248 + dd if=/dev/zero of=$DIR/$tfile bs=1M count=1 + cancel_lru_locks osc + sleep 1 + rm -f $TDIR/$tfile +} +run_test 819b "too big niobuf in write" + # # tests that do cleanup/setup should be run at the end #