X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Ftests%2Fconf-sanity.sh;h=34f2b3a711869c7ac9771bc6afa3c3b8b9308cdf;hb=f2a7347e4ceb31cefb43e3c61207fafe0a7af148;hp=13a6bfd2e8257908b3c8ee63109d45a9dc0579b3;hpb=9ab78a5bec154492c70f780706d9eaebeaaa5e54;p=fs%2Flustre-release.git diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh index 13a6bfd..34f2b3a 100644 --- a/lustre/tests/conf-sanity.sh +++ b/lustre/tests/conf-sanity.sh @@ -1,6 +1,4 @@ #!/bin/bash -# -*- mode: Bash; tab-width: 4; indent-tabs-mode: t; -*- -# vim:autoindent:shiftwidth=4:tabstop=4: # FIXME - there is no reason to use all of these different # return codes, espcially when most of them are mapped to something @@ -12,7 +10,6 @@ set -e ONLY=${ONLY:-"$*"} # bug number for skipped test: -# 15977 ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT" # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! @@ -22,6 +19,10 @@ if [ "$FAILURE_MODE" = "HARD" ]; then ALWAYS_EXCEPT="$ALWAYS_EXCEPT $CONFIG_EXCEPTIONS" fi +# LU-2059 +ALWAYS_EXCEPT="$ALWAYS_EXCEPT 5d 19b 21b 27a" + + SRCDIR=`dirname $0` PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH @@ -29,6 +30,7 @@ PTLDEBUG=${PTLDEBUG:--1} SAVE_PWD=$PWD LUSTRE=${LUSTRE:-`dirname $0`/..} RLUSTRE=${RLUSTRE:-$LUSTRE} +export MULTIOP=${MULTIOP:-multiop} . $LUSTRE/tests/test-framework.sh init_test_env $@ @@ -49,6 +51,11 @@ if [ -n "$MDSSIZE" ]; then STORED_MDSSIZE=$MDSSIZE fi +# pass "-E lazy_itable_init" to mke2fs to speed up the formatting time +if [[ "$LDISKFS_MKFS_OPTS" != *lazy_itable_init* ]]; then + LDISKFS_MKFS_OPTS=$(csa_add "$LDISKFS_MKFS_OPTS" -E lazy_itable_init) +fi + init_logging # @@ -190,6 +197,7 @@ setup() { start_mds || error "MDT start failed" start_ost || error "OST start failed" mount_client $MOUNT || error "client start failed" + client_up || error "client_up failed" } setup_noconfig() { @@ -215,7 +223,7 @@ cleanup_nocli() { } cleanup() { - umount_client $MOUNT || return 200 + umount_client $MOUNT || return 200 cleanup_nocli || return $? } @@ -316,7 +324,7 @@ test_5a() { # was test_5 # cleanup may return an error from the failed # disconnects; for now I'll consider this successful # if all the modules have unloaded. - umount -d $MOUNT & + umount -d $MOUNT & UMOUNT_PID=$! sleep 6 echo "killing umount" @@ -347,7 +355,7 @@ test_5a() { # was test_5 while [ "$WAIT" -ne "$MAX_WAIT" ]; do sleep $sleep grep -q $MOUNT" " /etc/mtab || break - echo "Waiting /etc/mtab updated ... " + echo "Waiting /etc/mtab updated ... " WAIT=$(( WAIT + sleep)) done [ "$WAIT" -eq "$MAX_WAIT" ] && error "/etc/mtab is not updated in $WAIT secs" @@ -554,6 +562,11 @@ is_blkdev () { # test_17() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + setup check_mount || return 41 cleanup || return $? @@ -576,7 +589,10 @@ test_17() { run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)" test_18() { - [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) @@ -610,15 +626,9 @@ test_18() { echo "mount mds with large journal..." - local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS - - local opts="--mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT" - if combined_mgs_mds ; then - MDS_MKFS_OPTS="--mgs $opts" - else - MDS_MKFS_OPTS="--mgsnode=$MGSNID $opts" - fi + local OLD_MDSSIZE=$MDSSIZE + MDSSIZE=$myMDSSIZE reformat_and_config echo "mount lustre system..." @@ -635,8 +645,8 @@ test_18() { cleanup || return $? - MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS - reformat_and_config + MDSSIZE=$OLD_MDSSIZE + reformat_and_config } run_test 18 "check mkfs creates large journals" @@ -722,6 +732,9 @@ test_21d() { stop_ost2 stop_mds stop_mgs + #writeconf to remove all ost2 traces for subsequent tests + writeconf + start_mgs } run_test 21d "start mgs then ost and then mds" @@ -751,6 +764,8 @@ test_22() { sleep $((TIMEOUT + TIMEOUT + TIMEOUT)) fi mount_client $MOUNT + wait_osc_import_state mds ost FULL + wait_osc_import_state client ost FULL check_mount || return 41 pass @@ -764,10 +779,10 @@ test_23a() { # was test_23 stop $SINGLEMDS # force down client so that recovering mds waits for reconnect local running=$(grep -c $MOUNT /proc/mounts) || true - if [ $running -ne 0 ]; then - echo "Stopping client $MOUNT (opts: -f)" - umount -f $MOUNT - fi + if [ $running -ne 0 ]; then + echo "Stopping client $MOUNT (opts: -f)" + umount -f $MOUNT + fi # enter recovery on mds start_mds @@ -779,10 +794,12 @@ test_23a() { # was test_23 echo mount pid is ${MOUNT_PID}, mount.lustre pid is ${MOUNT_LUSTRE_PID} ps --ppid $MOUNT_PID ps --ppid $MOUNT_LUSTRE_PID - # FIXME why o why can't I kill these? Manual "ctrl-c" works... - kill -TERM $MOUNT_LUSTRE_PID echo "waiting for mount to finish" ps -ef | grep mount + # "ctrl-c" sends SIGINT but it usually (in script) does not work on child process + # SIGTERM works but it does not spread to offspring processses + kill -s TERM $MOUNT_PID + kill -s TERM $MOUNT_LUSTRE_PID # we can not wait $MOUNT_PID because it is not a child of this shell local PID1 local PID2 @@ -825,6 +842,10 @@ run_test 23b "Simulate -EINTR during mount" fs2mds_HOST=$mds_HOST fs2ost_HOST=$ost_HOST +MDSDEV1_2=$fs2mds_DEV +OSTDEV1_2=$fs2ost_DEV +OSTDEV2_2=$fs3ost_DEV + cleanup_24a() { trap 0 echo "umount $MOUNT2 ..." @@ -845,14 +866,18 @@ test_24a() { [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST - local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} - local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2} + local fs2mdsdev=$(mdsdevname 1_2) + local fs2ostdev=$(ostdevname 1_2) + local fs2mdsvdev=$(mdsvdevname 1_2) + local fs2ostvdev=$(ostvdevname 1_2) # test 8-char fsname as well local FSNAME2=test1234 - add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --nomgs --mgsnode=$MGSNID --reformat $fs2mdsdev || exit 10 + add fs2mds $(mkfs_opts mds1) --nomgs --mgsnode=$MGSNID \ + --fsname=${FSNAME2} --reformat $fs2mdsdev $fs2mdsvdev || exit 10 - add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2ostdev || exit 10 + add fs2ost $(mkfs_opts ost1) --fsname=${FSNAME2} --reformat \ + $fs2ostdev $fs2ostvdev || exit 10 setup start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT @@ -875,7 +900,7 @@ test_24a() { facet_failover fs2mds facet_failover fs2ost df - umount_client $MOUNT + umount_client $MOUNT # the MDS must remain up until last MDT stop_mds MDS=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1) @@ -895,9 +920,11 @@ test_24b() { skip_env "mixed loopback and real device not working" && return fi - local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} + local fs2mdsdev=$(mdsdevname 1_2) + local fs2mdsvdev=$(mdsvdevname 1_2) - add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10 + add fs2mds $(mkfs_opts mds1) --mgs --fsname=${FSNAME}2 --reformat \ + $fs2mdsdev $fs2mdsvdev || exit 10 setup start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2 cleanup || return 6 @@ -923,33 +950,21 @@ test_26() { lctl get_param -n devices DEVS=$(lctl get_param -n devices | egrep -v MG | wc -l) [ $DEVS -gt 0 ] && return 2 + # start mds to drop writeconf setting + start_mds || return 3 + stop_mds || return 4 unload_modules_conf || return $? } run_test 26 "MDT startup failure cleans LOV (should return errs)" -set_and_check() { - local myfacet=$1 - local TEST=$2 - local PARAM=$3 - local ORIG=$(do_facet $myfacet "$TEST") - if [ $# -gt 3 ]; then - local FINAL=$4 - else - local -i FINAL - FINAL=$(($ORIG + 5)) - fi - echo "Setting $PARAM from $ORIG to $FINAL" - do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed - - wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed! -} - test_27a() { start_ost || return 1 start_mds || return 2 echo "Requeue thread should have started: " ps -e | grep ll_cfg_requeue - set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3 + set_conf_param_and_check ost1 \ + "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" \ + "$FSNAME-OST0000.ost.client_cache_seconds" || return 3 cleanup_nocli } run_test 27a "Reacquire MGS lock if OST started first" @@ -960,8 +975,12 @@ test_27b() { local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }') facet_failover $SINGLEMDS - set_and_check $SINGLEMDS "lctl get_param -n mdt.$device.identity_acquire_expire" "$device.mdt.identity_acquire_expire" || return 3 - set_and_check client "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight" "$device.mdc.max_rpcs_in_flight" || return 4 + set_conf_param_and_check $SINGLEMDS \ + "lctl get_param -n mdt.$device.identity_acquire_expire" \ + "$device.mdt.identity_acquire_expire" || return 3 + set_conf_param_and_check client \ + "lctl get_param -n mdc.$device-mdc-*.max_rpcs_in_flight"\ + "$device.mdc.max_rpcs_in_flight" || return 4 check_mount cleanup } @@ -973,10 +992,10 @@ test_28() { PARAM="$FSNAME.llite.max_read_ahead_whole_mb" ORIG=$($TEST) FINAL=$(($ORIG + 1)) - set_and_check client "$TEST" "$PARAM" $FINAL || return 3 + set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 3 FINAL=$(($FINAL + 1)) - set_and_check client "$TEST" "$PARAM" $FINAL || return 4 - umount_client $MOUNT || return 200 + set_conf_param_and_check client "$TEST" "$PARAM" $FINAL || return 4 + umount_client $MOUNT || return 200 mount_client $MOUNT RESULT=$($TEST) if [ $RESULT -ne $FINAL ]; then @@ -985,7 +1004,7 @@ test_28() { else echo "New config success: got $RESULT" fi - set_and_check client "$TEST" "$PARAM" $ORIG || return 5 + set_conf_param_and_check client "$TEST" "$PARAM" $ORIG || return 5 cleanup } run_test 28 "permanent parameter setting" @@ -1002,7 +1021,8 @@ test_29() { ACTV=$(lctl get_param -n $PROC_ACT) DEAC=$((1 - $ACTV)) - set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2 + set_conf_param_and_check client \ + "lctl get_param -n $PROC_ACT" "$PARAM" $DEAC || return 2 # also check ost_server_uuid status RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV) if [ -z "$RESULT" ]; then @@ -1034,11 +1054,8 @@ test_29() { echo "Waiting $(($MAX - $WAIT)) secs for MDT deactivated" done - # quotacheck should not fail immediately after deactivate - [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; } - # test new client starts deactivated - umount_client $MOUNT || return 200 + umount_client $MOUNT || return 200 mount_client $MOUNT RESULT=$(lctl get_param -n $PROC_UUID | grep DEACTIV | grep NEW) if [ -z "$RESULT" ]; then @@ -1048,13 +1065,11 @@ test_29() { echo "New client success: got $RESULT" fi - # quotacheck should not fail after umount/mount operation - [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; } - # make sure it reactivates - set_and_check client "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6 + set_conf_param_and_check client \ + "lctl get_param -n $PROC_ACT" "$PARAM" $ACTV || return 6 - umount_client $MOUNT + umount_client $MOUNT stop_ost2 cleanup_nocli #writeconf to remove all ost2 traces for subsequent tests @@ -1070,10 +1085,11 @@ test_30a() { ORIG=$($TEST) LIST=(1 2 3 4 5 4 3 2 1 2 3 4 5 4 3 2 1 2 3 4 5) for i in ${LIST[@]}; do - set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3 + set_conf_param_and_check client "$TEST" \ + "$FSNAME.llite.max_read_ahead_whole_mb" $i || return 3 done # make sure client restart still works - umount_client $MOUNT + umount_client $MOUNT mount_client $MOUNT || return 4 [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i" pass @@ -1104,7 +1120,9 @@ test_30b() { echo "Using fake nid $NEW" TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'" - set_and_check client "$TEST" "$FSNAME-OST0000.failover.node" $NEW || error "didn't add failover nid $NEW" + set_conf_param_and_check client "$TEST" \ + "$FSNAME-OST0000.failover.node" $NEW || + error "didn't add failover nid $NEW" NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids) echo $NIDS NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1)) @@ -1176,6 +1194,11 @@ cleanup_32() { } test_32a() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + client_only && skip "client only testing" && return 0 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; } [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0 @@ -1188,7 +1211,7 @@ test_32a() { { skip_env "Cannot untar $DISK1_8" && return 0; } load_modules - $LCTL set_param debug=$PTLDEBUG + $LCTL set_param debug="$PTLDEBUG" $TUNEFS $tmpdir/mds || error "tunefs failed" @@ -1243,6 +1266,11 @@ test_32a() { run_test 32a "Upgrade from 1.8 (not live)" test_32b() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + client_only && skip "client only testing" && return 0 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; } [ -z "$TUNEFS" ] && skip_env "No tunefs" && return @@ -1255,11 +1283,13 @@ test_32b() { { skip_env "Cannot untar $DISK1_8" && return ; } load_modules - $LCTL set_param debug="config" + $LCTL set_param debug="+config" local NEWNAME=lustre # writeconf will cause servers to register with their current nids - $TUNEFS --writeconf --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed" + $TUNEFS --writeconf --erase-params \ + --param mdt.identity_upcall=$L_GETIDENTITY \ + --fsname=$NEWNAME $tmpdir/mds || error "tunefs failed" combined_mgs_mds || stop mgs start32 mds1 $tmpdir/mds "-o loop" && \ @@ -1269,7 +1299,8 @@ test_32b() { echo MDS uuid $UUID [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID" - $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\ + $TUNEFS --writeconf --erase-params \ + --mgsnode=$HOSTNAME --fsname=$NEWNAME $tmpdir/ost1 ||\ error "tunefs failed" start32 ost1 $tmpdir/ost1 "-o loop" || return 5 UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid) @@ -1319,6 +1350,7 @@ test_33a() { # bug 12333, was test_33 local rc=0 local FSNAME2=test-123 local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) + local mkfsoptions [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST @@ -1329,12 +1361,19 @@ test_33a() { # bug 12333, was test_33 skip_env "mixed loopback and real device not working" && return fi - combined_mgs_mds || mkfs_opts="$mkfs_opts --nomgs" + local fs2mdsdev=$(mdsdevname 1_2) + local fs2ostdev=$(ostdevname 1_2) + local fs2mdsvdev=$(mdsvdevname 1_2) + local fs2ostvdev=$(ostvdevname 1_2) + + if [ $(facet_fstype mds1) == ldiskfs ]; then + mkfsoptions="--mkfsoptions=\\\"-J size=8\\\"" # See bug 17931. + fi - local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} - local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2} - add fs2mds $MDS_MKFS_OPTS --mkfsoptions='\"-J size=8\"' --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10 - add fs2ost $OST_MKFS_OPTS --fsname=${FSNAME2} --index=8191 --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10 + add fs2mds $(mkfs_opts mds1) --fsname=${FSNAME2} --reformat \ + $mkfsoptions $fs2mdsdev $fs2mdsvdev || exit 10 + add fs2ost $(mkfs_opts ost1) --mgsnode=$MGSNID --fsname=${FSNAME2} \ + --index=8191 --reformat $fs2ostdev $fs2ostvdev || exit 10 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT start fs2ost $fs2ostdev $OST_MOUNT_OPTS @@ -1349,7 +1388,6 @@ test_33a() { # bug 12333, was test_33 umount -d $MOUNT2 stop fs2ost -f stop fs2mds -f - rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev cleanup_nocli || rc=6 return $rc } @@ -1388,7 +1426,7 @@ test_34b() { touch $DIR/$tfile || return 1 stop_mds --force || return 2 - manual_umount_client --force + manual_umount_client --force rc=$? if [ $rc -ne 0 ]; then error "mtab after failed umount - rc $rc" @@ -1404,7 +1442,7 @@ test_34c() { touch $DIR/$tfile || return 1 stop_ost --force || return 2 - manual_umount_client --force + manual_umount_client --force rc=$? if [ $rc -ne 0 ]; then error "mtab after failed umount - rc $rc" @@ -1424,7 +1462,7 @@ test_35a() { # bug 12459 log "Set up a fake failnode for the MDS" FAKENID="127.0.0.2" local device=$(do_facet $SINGLEMDS "lctl get_param -n devices" | awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1) - do_facet $SINGLEMDS $LCTL conf_param ${device}.failover.node=$FAKENID || return 4 + do_facet mgs $LCTL conf_param ${device}.failover.node=$FAKENID || return 4 log "Wait for RECONNECT_INTERVAL seconds (10s)" sleep 10 @@ -1478,29 +1516,31 @@ test_35b() { # bug 18674 FAKENID="127.0.0.2" local device=$(do_facet $SINGLEMDS "$LCTL get_param -n devices" | \ awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1) - do_facet $SINGLEMDS "$LCTL conf_param ${device}.failover.node=$FAKENID" || \ + do_facet mgs "$LCTL conf_param ${device}.failover.node=$FAKENID" || \ return 1 local at_max_saved=0 - # adaptive timeouts may prevent seeing the issue + # adaptive timeouts may prevent seeing the issue if at_is_enabled; then at_max_saved=$(at_max_get mds) at_max_set 0 mds client fi - mkdir -p $MOUNT/testdir - touch $MOUNT/testdir/test + mkdir -p $MOUNT/$tdir log "Injecting EBUSY on MDS" # Setting OBD_FAIL_MDS_RESEND=0x136 do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000136" || return 2 - log "Stat on a test file" - stat $MOUNT/testdir/test + $LCTL set_param mdc.${FSNAME}*.stats=clear + + log "Creating a test file and stat it" + touch $MOUNT/$tdir/$tfile + stat $MOUNT/$tdir/$tfile log "Stop injecting EBUSY on MDS" do_facet $SINGLEMDS "$LCTL set_param fail_loc=0" || return 3 - rm -f $MOUNT/testdir/test + rm -f $MOUNT/$tdir/$tfile log "done" # restore adaptive timeout @@ -1508,6 +1548,8 @@ test_35b() { # bug 18674 $LCTL dk $TMP/lustre-log-$TESTNAME.log + CONNCNT=`$LCTL get_param mdc.${FSNAME}*.stats | awk '/mds_connect/{print $2}'` + # retrieve from the log if the client has ever tried to # contact the fake server after the loss of connection FAILCONN=`awk "BEGIN {ret = 0;} @@ -1527,7 +1569,14 @@ test_35b() { # bug 18674 log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \ return 5 - cleanup + # LU-290 + # When OBD_FAIL_MDS_RESEND is hit, we sleep for 2 * obd_timeout + # Reconnects are supposed to be rate limited to one every 5s + [ $CONNCNT -gt $((2 * $TIMEOUT / 5 + 1)) ] && \ + log "ERROR: Too many reconnects $CONNCNT" && \ + return 6 + + cleanup # remove nid settings writeconf } @@ -1551,14 +1600,21 @@ test_36() { # 12743 skip_env "mixed loopback and real device not working" && return fi - local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} - local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2} - local fs3ostdev=${fs3ost_DEV:-$(ostdevname 2)_2} - add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10 - # XXX after we support non 4K disk blocksize, change following --mkfsoptions with - # other argument - add fs2ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs2ostdev || exit 10 - add fs3ost $OST_MKFS_OPTS --mkfsoptions='-b4096' --fsname=${FSNAME2} --mgsnode=$MGSNID --reformat $fs3ostdev || exit 10 + local fs2mdsdev=$(mdsdevname 1_2) + local fs2ostdev=$(ostdevname 1_2) + local fs3ostdev=$(ostdevname 2_2) + local fs2mdsvdev=$(mdsvdevname 1_2) + local fs2ostvdev=$(ostvdevname 1_2) + local fs3ostvdev=$(ostvdevname 2_2) + + add fs2mds $(mkfs_opts mds1) --fsname=${FSNAME2} --reformat \ + $fs2mdsdev $fs2mdsvdev || exit 10 + # XXX after we support non 4K disk blocksize in ldiskfs, specify a + # different one than the default value here. + add fs2ost $(mkfs_opts ost1) --mgsnode=$MGSNID --fsname=${FSNAME2} \ + --reformat $fs2ostdev $fs2ostvdev || exit 10 + add fs3ost $(mkfs_opts ost1) --mgsnode=$MGSNID --fsname=${FSNAME2} \ + --reformat $fs3ostdev $fs3ostvdev || exit 10 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS start fs2ost $fs2ostdev $OST_MOUNT_OPTS @@ -1601,7 +1657,6 @@ test_36() { # 12743 stop fs3ost -f || return 200 stop fs2ost -f || return 201 stop fs2mds -f || return 202 - rm -rf $MOUNT2 $fs2mdsdev $fs2ostdev $fs3ostdev unload_modules_conf || return 203 return $rc } @@ -1611,6 +1666,13 @@ test_37() { local mntpt=$(facet_mntpt $SINGLEMDS) local mdsdev=$(mdsdevname ${SINGLEMDS//mds/}) local mdsdev_sym="$TMP/sym_mdt.img" + local opts=$MDS_MOUNT_OPTS + local rc=0 + + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Currently only applicable to ldiskfs-based MDTs" + return + fi echo "MDS : $mdsdev" echo "SYMLINK : $mdsdev_sym" @@ -1620,8 +1682,11 @@ test_37() { echo "mount symlink device - $mdsdev_sym" - local rc=0 - mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $MDS_MOUNT_OPTS $mdsdev_sym $mntpt 2>&1 ) + if ! do_facet $SINGLEMDS test -b $mdsdev; then + opts=$(csa_add "$opts" -o loop) + fi + mount_op=$(do_facet $SINGLEMDS mount -v -t lustre $opts \ + $mdsdev_sym $mntpt 2>&1) rc=${PIPESTATUS[0]} echo mount_op=$mount_op @@ -1639,6 +1704,11 @@ test_37() { run_test 37 "verify set tunables works for symlink device" test_38() { # bug 14222 + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + setup # like runtests COUNT=10 @@ -1707,14 +1777,18 @@ run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages" test_40() { # bug 15759 start_ost #define OBD_FAIL_TGT_TOOMANY_THREADS 0x706 - do_facet $SINGLEMDS "sysctl -w lustre.fail_loc=0x80000706" + do_facet $SINGLEMDS "$LCTL set_param fail_loc=0x80000706" start_mds cleanup } run_test 40 "race during service thread startup" test_41a() { #bug 14134 - echo $MDS_MOUNT_OPTS | grep "loop" && skip " loop devices does not work with nosvc option" && return + if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] && + ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then + skip "Loop devices does not work with nosvc option" + return + fi local rc local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) @@ -1739,7 +1813,13 @@ test_41a() { #bug 14134 run_test 41a "mount mds with --nosvc and --nomgs" test_41b() { - echo $MDS_MOUNT_OPTS | grep "loop" && skip " loop devices does not work with nosvc option" && return + if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] && + ! do_facet $SINGLEMDS test -b $(mdsdevname 1); then + skip "Loop devices does not work with nosvc option" + return + fi + + ! combined_mgs_mds && skip "needs combined mgs device" && return 0 stopall reformat @@ -1761,8 +1841,8 @@ test_41b() { stop_mds -f || return 203 } - run_test 41b "mount mds with --nosvc and --nomgs on first mount" + test_42() { #bug 14693 setup check_mount || return 2 @@ -1775,17 +1855,17 @@ test_42() { #bug 14693 run_test 42 "invalid config param should not prevent client from mounting" test_43() { - [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root" - setup - chmod ugo+x $DIR || error "chmod 0 failed" - set_and_check mds \ - "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \ - "$FSNAME.mdt.root_squash" \ - "0:0" - set_and_check mds \ - "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \ - "$FSNAME.mdt.nosquash_nids" \ - "NONE" + [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root" + setup + chmod ugo+x $DIR || error "chmod 0 failed" + set_conf_param_and_check mds \ + "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \ + "$FSNAME.mdt.root_squash" \ + "0:0" + set_conf_param_and_check mds \ + "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \ + "$FSNAME.mdt.nosquash_nids" \ + "NONE" # # create set of test files @@ -1801,15 +1881,15 @@ test_43() { chmod go-rwx $DIR/$tdir-rootdir || error "chmod 3 failed" touch $DIR/$tdir-rootdir/tfile-1 || error "touch failed" - # - # check root_squash: - # set root squash UID:GID to RUNAS_ID - # root should be able to access only files owned by RUNAS_ID - # - set_and_check mds \ - "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \ - "$FSNAME.mdt.root_squash" \ - "$RUNAS_ID:$RUNAS_ID" + # + # check root_squash: + # set root squash UID:GID to RUNAS_ID + # root should be able to access only files owned by RUNAS_ID + # + set_conf_param_and_check mds \ + "lctl get_param -n mdt.$FSNAME-MDT0000.root_squash" \ + "$FSNAME.mdt.root_squash" \ + "$RUNAS_ID:$RUNAS_ID" ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile) dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \ @@ -1840,18 +1920,18 @@ test_43() { error "$ST: root create permission is granted" echo "$ST: root create permission is denied - ok" - # - # check nosquash_nids: - # put client's NID into nosquash_nids list, - # root should be able to access root file after that - # - local NIDLIST=$(lctl list_nids all | tr '\n' ' ') - NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp" - NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ') - set_and_check mds \ - "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \ - "$FSNAME-MDTall.mdt.nosquash_nids" \ - "$NIDLIST" + # + # check nosquash_nids: + # put client's NID into nosquash_nids list, + # root should be able to access root file after that + # + local NIDLIST=$(lctl list_nids all | tr '\n' ' ') + NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp" + NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ') + set_conf_param_and_check mds \ + "lctl get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \ + "$FSNAME-MDTall.mdt.nosquash_nids" \ + "$NIDLIST" ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile) dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \ @@ -1926,7 +2006,7 @@ cleanup_46a() { stop ost${count} -f || rc=$? let count=count-1 done - stop_mds || rc=$? + stop_mds || rc=$? cleanup_nocli || rc=$? #writeconf to remove all ost2 traces for subsequent tests writeconf @@ -1944,7 +2024,7 @@ test_46a() { mount_client $MOUNT || return 3 trap "cleanup_46a $OSTCOUNT" EXIT ERR - local i + local i for (( i=2; i<=$OSTCOUNT; i++ )); do start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2)) done @@ -1952,13 +2032,13 @@ test_46a() { # wait until osts in sync for (( i=2; i<=$OSTCOUNT; i++ )); do wait_osc_import_state mds ost$i FULL + wait_osc_import_state client ost$i FULL done - #second client see all ost's mount_client $MOUNT2 || return 8 - $LFS setstripe $MOUNT2 -c -1 || return 9 + $LFS setstripe -c -1 $MOUNT2 || return 9 $LFS getstripe $MOUNT2 || return 10 echo "ok" > $MOUNT2/widestripe @@ -2026,9 +2106,9 @@ cleanup_48() { test_48() { # bug 17636 reformat setup_noconfig - check_mount || return 2 + check_mount || return 2 - $LFS setstripe $MOUNT -c -1 || return 9 + $LFS setstripe -c -1 $MOUNT || return 9 $LFS getstripe $MOUNT || return 10 echo "ok" > $MOUNT/widestripe @@ -2050,12 +2130,12 @@ run_test 48 "too many acls on file" # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE test_49() { # bug 17710 - local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS - local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS + local timeout_orig=$TIMEOUT + local ldlm_timeout_orig=$LDLM_TIMEOUT local LOCAL_TIMEOUT=20 - - OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT" + LDLM_TIMEOUT=$LOCAL_TIMEOUT + TIMEOUT=$LOCAL_TIMEOUT reformat setup_noconfig @@ -2078,7 +2158,7 @@ test_49() { # bug 17710 stop_ost || return 2 stop_mds || return 3 - OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT" + LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1)) reformat setup_noconfig @@ -2098,8 +2178,8 @@ test_49() { # bug 17710 cleanup || return $? - MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS - OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS + LDLM_TIMEOUT=$ldlm_timeout_orig + TIMEOUT=$timeout_orig } run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE" @@ -2167,9 +2247,11 @@ test_50c() { wait_osc_import_state mds ost DISCONN lazystatfs $MOUNT || error "lazystatfs failed with one down server" - umount_client $MOUNT || error "Unable to unmount client" + umount_client $MOUNT || error "Unable to unmount client" stop_ost2 || error "Unable to stop OST2" stop_mds || error "Unable to stop MDS" + #writeconf to remove all ost2 traces for subsequent tests + writeconf } run_test 50c "lazystatfs one server down ==========================" @@ -2187,9 +2269,11 @@ test_50d() { stop_ost || error "Unable to stop OST1" lazystatfs $MOUNT || error "lazystatfs failed with one down server" - umount_client $MOUNT || error "Unable to unmount client" + umount_client $MOUNT || error "Unable to unmount client" stop_ost2 || error "Unable to stop OST2" stop_mds || error "Unable to stop MDS" + #writeconf to remove all ost2 traces for subsequent tests + writeconf } run_test 50d "lazystatfs client/server conn race ==========================" @@ -2321,6 +2405,8 @@ test_51() { wait $pid stop_ost2 || return 3 cleanup + #writeconf to remove all ost2 traces for subsequent tests + writeconf } run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added" @@ -2373,6 +2459,11 @@ diff_files_xattrs() } test_52() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + start_mds [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; } start_ost @@ -2381,9 +2472,10 @@ test_52() { [ $? -eq 0 ] || { error "Unable to mount client"; return 3; } local nrfiles=8 - local ost1mnt=${MOUNT%/*}/ost1 + local ost1mnt=$(facet_mntpt ost1) local ost1node=$(facet_active_host ost1) local ost1tmp=$TMP/conf52 + local loop mkdir -p $DIR/$tdir [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; } @@ -2395,7 +2487,7 @@ test_52() { [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; } sleep 1 - $LFS setstripe $DIR/$tdir -c -1 -s 1M + $LFS setstripe -c -1 -S 1M $DIR/$tdir [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; } for (( i=0; i < nrfiles; i++ )); do @@ -2417,7 +2509,13 @@ test_52() { [ $? -eq 0 ] || { error "Unable to stop ost1"; return 11; } echo mount ost1 as ldiskfs - do_node $ost1node mount -t $FSTYPE $ost1_dev $ost1mnt $OST_MOUNT_OPTS + do_node $ost1node mkdir -p $ost1mnt + [ $? -eq 0 ] || { error "Unable to create $ost1mnt"; return 23; } + if ! do_node $ost1node test -b $ost1_dev; then + loop="-o loop" + fi + do_node $ost1node mount -t $(facet_fstype ost1) $loop $ost1_dev \ + $ost1mnt [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; } # backup objects @@ -2439,7 +2537,7 @@ test_52() { diff_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects [ $? -eq 0 ] || { error "Unable to diff objects"; return 16; } - do_node $ost1node "umount $ost1_dev" + do_node $ost1node "umount $ost1mnt" [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; } start_ost @@ -2468,6 +2566,7 @@ thread_sanity() { local facet=$2 local parampat=$3 local opts=$4 + local basethr=$5 local tmin local tmin2 local tmax @@ -2475,6 +2574,8 @@ thread_sanity() { local tstarted local paramp local msg="Insane $modname thread counts" + local ncpts=$(check_cpt_number $facet) + local nthrs shift 4 setup @@ -2496,13 +2597,23 @@ thread_sanity() { tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started" || echo 0) lassert 23 "$msg (PDSH problems?)" '(($tstarted && $tmin && $tmax))' || return $? lassert 24 "$msg" '(($tstarted >= $tmin && $tstarted <= $tmax ))' || return $? + nthrs=$(expr $tmax - $tmin) + if [ $nthrs -lt $ncpts ]; then + nthrs=0 + else + nthrs=$ncpts + fi + + [ $tmin -eq $tmax -a $tmin -eq $tstarted ] && + skip_env "module parameter forced $facet thread count" && + tmin=3 && tmax=$((3 * tmax)) # Check that we can change min/max - do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + 1))" - do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 1))" - tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0) - tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0) - lassert 25 "$msg" '(($tmin2 == ($tmin + 1) && $tmax2 == ($tmax -1)))' || return $? + do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin + nthrs))" + do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - nthrs))" + tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0) + tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0) + lassert 25 "$msg" '(($tmin2 == ($tmin + $nthrs) && $tmax2 == ($tmax - $nthrs)))' || return $? # Check that we can set min/max to the same value tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0) @@ -2522,7 +2633,8 @@ thread_sanity() { LOAD_MODULES_REMOTE=true cleanup local oldvalue - setmodopts -a $modname "$opts" oldvalue + local newvalue="${opts}=$(expr $basethr \* $ncpts)" + setmodopts -a $modname "$newvalue" oldvalue load_modules setup @@ -2538,51 +2650,45 @@ thread_sanity() { lassert 28 "$msg" '(($tstarted == $tmin && $tstarted == $tmax ))' || return $? cleanup - # Workaround a YALA bug where YALA expects that modules will remain - # loaded on the servers - LOAD_MODULES_REMOTE=false load_modules setup cleanup } test_53a() { - thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads=64' + thread_sanity OST ost1 'ost.*.ost' 'oss_num_threads' '16' } run_test 53a "check OSS thread count params" test_53b() { - thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64' + thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads' '16' } run_test 53b "check MDT thread count params" -run_llverfs() -{ - local dir=$1 - local partial_arg="" - local size=$(df -B G $dir | tail -1 | awk '{print $2}' | sed 's/G//') # Gb - - # Run in partial (fast) mode if the size - # of a partition > 10 GB - [ $size -gt 10 ] && partial_arg="-p" - - llverfs $partial_arg $dir -} - test_54a() { - do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + + do_rpc_nodes $(facet_host ost1) run_llverdev $(ostdevname 1) -p [ $? -eq 0 ] || error "llverdev failed!" reformat_and_config } -run_test 54a "llverdev" +run_test 54a "test llverdev and partial verify of device" test_54b() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + setup - run_llverfs $MOUNT + run_llverfs $MOUNT -p [ $? -eq 0 ] || error "llverfs failed!" cleanup } -run_test 54b "llverfs" +run_test 54b "test llverfs and partial verify of filesystem" lov_objid_size() { @@ -2591,20 +2697,25 @@ lov_objid_size() } test_55() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + local mdsdev=$(mdsdevname 1) - local ostdev=$(ostdevname 1) - local saved_opts=$OST_MKFS_OPTS + local mdsvdev=$(mdsvdevname 1) - for i in 0 1023 2048 + for i in 1023 2048 do - OST_MKFS_OPTS="$saved_opts --index $i" - reformat - + add mds1 $(mkfs_opts mds1) --reformat $mdsdev $mdsvdev || + exit 10 + add ost1 $(mkfs_opts ost1) --index=$i --reformat \ + $(ostdevname 1) $(ostvdevname 1) setup_noconfig stopall - - setup + setup_noconfig sync + echo checking size of lov_objid for ost index $i LOV_OBJID_SIZE=$(do_facet mds1 "$DEBUGFS -R 'stat lov_objid' $mdsdev 2>/dev/null" | grep ^User | awk '{print $6}') if [ "$LOV_OBJID_SIZE" != $(lov_objid_size $i) ]; then @@ -2615,43 +2726,81 @@ test_55() { stopall done - OST_MKFS_OPTS=$saved_opts reformat } run_test 55 "check lov_objid size" test_56() { - add mds1 $MDS_MKFS_OPTS --mkfsoptions='\"-J size=16\"' --reformat $(mdsdevname 1) - add ost1 $OST_MKFS_OPTS --index=1000 --reformat $(ostdevname 1) - add ost2 $OST_MKFS_OPTS --index=10000 --reformat $(ostdevname 2) + local mds_journal_size_orig=$MDSJOURNALSIZE - start_mds + MDSJOURNALSIZE=16 + add mds1 $(mkfs_opts mds1) --reformat $(mdsdevname 1) $(mdsvdevname 1) + add ost1 $(mkfs_opts ost1) --index=1000 --reformat \ + $(ostdevname 1) $(ostvdevname 1) + add ost2 $(mkfs_opts ost2) --index=10000 --reformat \ + $(ostdevname 2) $(ostvdevname 2) + + start_mgsmds start_ost start_ost2 || error "Unable to start second ost" mount_client $MOUNT || error "Unable to mount client" echo ok $LFS osts - [ -n "$ENABLE_QUOTA" ] && { $LFS quotacheck -ug $MOUNT || error "quotacheck has failed" ; } stopall + MDSJOURNALSIZE=$mds_journal_size_orig reformat } run_test 56 "check big indexes" -test_57() { # bug 22656 +test_57a() { # bug 22656 local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}') writeconf do_facet ost1 "$TUNEFS --failnode=$NID `ostdevname 1`" || error "tunefs failed" start_mgsmds start_ost && error "OST registration from failnode should fail" - stop_mds reformat } -run_test 57 "initial registration from failnode should fail (should return errs)" +run_test 57a "initial registration from failnode should fail (should return errs)" + +test_57b() { + local NID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}') + writeconf + do_facet ost1 "$TUNEFS --servicenode=$NID `ostdevname 1`" || error "tunefs failed" + start_mgsmds + start_ost || error "OST registration from servicenode should not fail" + reformat +} +run_test 57b "initial registration from servicenode should not fail" count_osts() { do_facet mgs $LCTL get_param mgs.MGS.live.$FSNAME | grep OST | wc -l } +test_58() { # bug 22658 + if [ $(facet_fstype mds) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + setup_noconfig + mkdir -p $DIR/$tdir + createmany -o $DIR/$tdir/$tfile-%d 100 + # make sure that OSTs do not cancel llog cookies before we unmount the MDS +#define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601 + do_facet mds "lctl set_param fail_loc=0x601" + unlinkmany $DIR/$tdir/$tfile-%d 100 + stop mds + local MNTDIR=$(facet_mntpt mds) + # remove all files from the OBJECTS dir + do_facet mds "mount -t ldiskfs $MDSDEV $MNTDIR" + do_facet mds "find $MNTDIR/OBJECTS -type f -delete" + do_facet mds "umount $MNTDIR" + # restart MDS with missing llog files + start_mds + do_facet mds "lctl set_param fail_loc=0" + reformat +} +run_test 58 "missing llog files must not prevent MDT from mounting" + test_59() { start_mgsmds >> /dev/null local C1=$(count_osts) @@ -2679,31 +2828,157 @@ test_59() { [ $C4 -eq 2 ] || error "OST2 writeconf should add log" stop_ost2 >> /dev/null cleanup_nocli >> /dev/null + #writeconf to remove all ost2 traces for subsequent tests + writeconf } run_test 59 "writeconf mount option" +test_60() { # LU-471 + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi -test_58() { # bug 22658 - [ "$FSTYPE" != "ldiskfs" ] && skip "not supported for $FSTYPE" && return - setup - mkdir -p $DIR/$tdir - createmany -o $DIR/$tdir/$tfile-%d 100 - # make sure that OSTs do not cancel llog cookies before we unmount the MDS -#define OBD_FAIL_OBD_LOG_CANCEL_NET 0x601 - do_facet mds "lctl set_param fail_loc=0x601" - unlinkmany $DIR/$tdir/$tfile-%d 100 - stop mds - local MNTDIR=$(facet_mntpt mds) - # remove all files from the OBJECTS dir - do_facet mds "mount -t ldiskfs $MDSDEV $MNTDIR" - do_facet mds "find $MNTDIR/OBJECTS -type f -delete" - do_facet mds "umount $MNTDIR" - # restart MDS with missing llog files - start_mds - do_facet mds "lctl set_param fail_loc=0" + add mds1 $(mkfs_opts mds1) \ + --mkfsoptions='\" -E stride=64 -O ^uninit_bg\"' --reformat \ + $(mdsdevname 1) $(mdsvdevname 1) || exit 10 + + dump=$(do_facet $SINGLEMDS dumpe2fs $(mdsdevname 1)) + rc=${PIPESTATUS[0]} + [ $rc -eq 0 ] || error "dumpe2fs $(mdsdevname 1) failed" + + # MDT default has dirdata feature + echo $dump | grep dirdata > /dev/null || error "dirdata is not set" + # we disable uninit_bg feature + echo $dump | grep uninit_bg > /dev/null && error "uninit_bg is set" + # we set stride extended options + echo $dump | grep stride > /dev/null || error "stride is not set" reformat } -run_test 58 "missing llog files must not prevent MDT from mounting" +run_test 60 "check mkfs.lustre --mkfsoptions -E -O options setting" + +test_61() { # LU-80 + local reformat=false + + [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.1.53) ] || + { skip "Need MDS version at least 2.1.53"; return 0; } + + if [ $(facet_fstype $SINGLEMDS) == ldiskfs ] && + ! large_xattr_enabled; then + reformat=true + local mds_dev=$(mdsdevname ${SINGLEMDS//mds/}) + LDISKFS_MKFS_OPTS+=" -O large_xattr" + add $SINGLEMDS $(mkfs_opts $SINGLEMDS) --reformat $mds_dev || + error "reformatting $mds_dev failed" + fi + + setup_noconfig || error "setting up the filesystem failed" + client_up || error "starting client failed" + + local file=$DIR/$tfile + touch $file + + local large_value="$(generate_string $(max_xattr_size))" + local small_value="bar" + + local name="trusted.big" + log "save large xattr $name on $file" + setfattr -n $name -v $large_value $file || + error "saving $name on $file failed" + + local new_value=$(get_xattr_value $name $file) + [[ "$new_value" != "$large_value" ]] && + error "$name different after saving" + + log "shrink value of $name on $file" + setfattr -n $name -v $small_value $file || + error "shrinking value of $name on $file failed" + + new_value=$(get_xattr_value $name $file) + [[ "$new_value" != "$small_value" ]] && + error "$name different after shrinking" + + log "grow value of $name on $file" + setfattr -n $name -v $large_value $file || + error "growing value of $name on $file failed" + + new_value=$(get_xattr_value $name $file) + [[ "$new_value" != "$large_value" ]] && + error "$name different after growing" + + log "check value of $name on $file after remounting MDS" + fail $SINGLEMDS + new_value=$(get_xattr_value $name $file) + [[ "$new_value" != "$large_value" ]] && + error "$name different after remounting MDS" + + log "remove large xattr $name from $file" + setfattr -x $name $file || error "removing $name from $file failed" + + rm -f $file + stopall + if $reformat; then + LDISKFS_MKFS_OPTS=${LDISKFS_MKFS_OPTS% -O large_xattr} + reformat + fi +} +run_test 61 "large xattr" + +test_62() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + + # MRP-118 + local mdsdev=$(mdsdevname 1) + local ostdev=$(ostdevname 1) + + [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.2.51) ]] || + { skip "Need MDS version at least 2.2.51"; return 0; } + + echo "disable journal for mds" + do_facet mds tune2fs -O ^has_journal $mdsdev || error "tune2fs failed" + start_mds && error "MDT start should fail" + echo "disable journal for ost" + do_facet ost1 tune2fs -O ^has_journal $ostdev || error "tune2fs failed" + start_ost && error "OST start should fail" + cleanup || return $? + reformat_and_config +} +run_test 62 "start with disabled journal" + +test_63() { + if [ $(facet_fstype $SINGLEMDS) != ldiskfs ]; then + skip "Only applicable to ldiskfs-based MDTs" + return + fi + + local inode_slab=$(do_facet $SINGLEMDS \ + "awk '/ldiskfs_inode_cache/ { print \\\$5 }' /proc/slabinfo") + if [ -z "$inode_slab" ]; then + skip "ldiskfs module has not been loaded" + return + fi + + echo "$inode_slab ldisk inodes per page" + [ "$inode_slab" -ge "3" ] || + error "ldisk inode size is too big, $inode_slab objs per page" + return +} +run_test 63 "Verify each page can at least hold 3 ldisk inodes" + +test_64() { + start_mds + start_ost + start_ost2 || error "Unable to start second ost" + mount_client $MOUNT || error "Unable to mount client" + stop_ost2 || error "Unable to stop second ost" + echo "$LFS df" + $LFS df --lazy || error "lfs df failed" + cleanup || return $? +} +run_test 64 "check lfs df --lazy " if ! combined_mgs_mds ; then stop mgs