X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fconf-sanity.sh;h=d1bc9f4742a510dae426af09bb106032e4fe37b1;hp=7ab509ee4ef5ac796db3757e56821c5cde9fc8e1;hb=9229c0dec72e1da56df04749ab06fbb2e7713729;hpb=f433a28365fe0c50cbe70ad6700e2520918d2354 diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh index 7ab509e..d1bc9f4 100644 --- a/lustre/tests/conf-sanity.sh +++ b/lustre/tests/conf-sanity.sh @@ -1,6 +1,6 @@ #!/bin/bash -# requirement: -# add uml1 uml2 uml3 in your /etc/hosts +# -*- 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 @@ -11,11 +11,9 @@ set -e ONLY=${ONLY:-"$*"} -# bug number for skipped test: 13739 -HEAD_EXCEPT=" 32a" - -# bug number for skipped test: -ALWAYS_EXCEPT=" $CONF_SANITY_EXCEPT $HEAD_EXCEPT" +# bug number for skipped test: +# 15977 +ALWAYS_EXCEPT="$CONF_SANITY_EXCEPT" # UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT! SRCDIR=`dirname $0` @@ -28,19 +26,23 @@ RLUSTRE=${RLUSTRE:-$LUSTRE} . $LUSTRE/tests/test-framework.sh init_test_env $@ + +# use small MDS + OST size to speed formatting time +# do not use too small MDSSIZE/OSTSIZE, which affect the default jouranl size +MDSSIZE=200000 +OSTSIZE=200000 +. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh} + # STORED_MDSSIZE is used in test_18 if [ -n "$MDSSIZE" ]; then STORED_MDSSIZE=$MDSSIZE fi -# use small MDS + OST size to speed formatting time -MDSSIZE=40000 -OSTSIZE=40000 -. ${CONFIG:=$LUSTRE/tests/cfg/$NAME.sh} -remote_mds_nodsh && skip "remote MDS with nodsh" && exit 0 -remote_ost_nodsh && skip "remote OST with nodsh" && exit 0 +init_logging + +require_dsh_mds || exit 0 +require_dsh_ost || exit 0 -# [ "$SLOW" = "no" ] && EXCEPT_SLOW="0 1 2 3 6 7 15 18 24b 25 30 31 32 33 34a 45" assert_DIR @@ -50,26 +52,40 @@ reformat() { } writeconf() { - local facet=$SINGLEMDS - local dev=${facet}_dev - shift - stop ${facet} -f - rm -f ${facet}active - # who knows if/where $TUNEFS is installed? Better reformat if it fails... - do_facet ${facet} "$TUNEFS --writeconf ${!dev}" || echo "tunefs failed, reformatting instead" && reformat + local facet=$SINGLEMDS + local dev=${facet}_dev + shift + stop ${facet} -f + rm -f ${facet}active + # who knows if/where $TUNEFS is installed? Better reformat if it fails... + do_facet ${facet} "$TUNEFS --writeconf ${!dev}" || + echo "tunefs failed, reformatting instead" && reformat_and_config + } gen_config() { - reformat - # The MGS must be started before the OSTs for a new fs, so start - # and stop to generate the startup logs. + # The MGS must be started before the OSTs for a new fs, so start + # and stop to generate the startup logs. start_mds start_ost - sleep 5 + wait_osc_import_state mds ost FULL stop_ost stop_mds } +reformat_and_config() { + reformat + if ! combined_mgs_mds ; then + start_mgs + fi + gen_config +} + +start_mgs () { + echo "start mgs" + start mgs $MGSDEV $mgs_MOUNT_OPTS +} + start_mds() { local facet=$SINGLEMDS # we can not use MDSDEV1 here because SINGLEMDS could be set not to mds1 only @@ -138,20 +154,24 @@ manual_umount_client(){ } setup() { - start_ost - start_mds - mount_client $MOUNT + start_mds || error "MDT start failed" + start_ost || error "OST start failed" + mount_client $MOUNT || error "client start failed" } setup_noconfig() { + if ! combined_mgs_mds ; then + start_mgs + fi + start_mds start_ost mount_client $MOUNT } cleanup_nocli() { - stop_mds || return 201 stop_ost || return 202 + stop_mds || return 201 unload_modules || return 203 } @@ -163,17 +183,17 @@ cleanup() { check_mount() { do_facet client "cp /etc/passwd $DIR/a" || return 71 do_facet client "rm $DIR/a" || return 72 - # make sure lustre is actually mounted (touch will block, - # but grep won't, so do it after) + # make sure lustre is actually mounted (touch will block, + # but grep won't, so do it after) do_facet client "grep $MOUNT' ' /proc/mounts > /dev/null" || return 73 echo "setup single mount lustre success" } check_mount2() { - do_facet client "touch $DIR/a" || return 71 - do_facet client "rm $DIR/a" || return 72 - do_facet client "touch $DIR2/a" || return 73 - do_facet client "rm $DIR2/a" || return 74 + do_facet client "touch $DIR/a" || return 71 + do_facet client "rm $DIR/a" || return 72 + do_facet client "touch $DIR2/a" || return 73 + do_facet client "rm $DIR2/a" || return 74 echo "setup double mount lustre success" } @@ -193,7 +213,7 @@ init_gss #create single point mountpoint -gen_config +reformat_and_config test_0() { setup @@ -203,19 +223,21 @@ test_0() { run_test 0 "single mount setup" test_1() { + start_mds || error "MDT start failed" start_ost echo "start ost second time..." - setup + start_ost && error "2nd OST start should fail" + mount_client $MOUNT || error "client start failed" check_mount || return 42 cleanup || return $? } run_test 1 "start up ost twice (should return errors)" test_2() { - start_ost start_mds echo "start mds second time.." - start_mds + start_mds && error "2nd MDT start should fail" + start_ost mount_client $MOUNT check_mount || return 43 cleanup || return $? @@ -225,7 +247,7 @@ run_test 2 "start up mds twice (should return err)" test_3() { setup #mount.lustre returns an error if already in mtab - mount_client $MOUNT && return $? + mount_client $MOUNT && error "2nd client mount should fail" check_mount || return 44 cleanup || return $? } @@ -308,8 +330,8 @@ test_5b() { run_test 5b "mds down, cleanup after failed mount (bug 2712) (should return errs)" test_5c() { - start_ost start_mds + start_ost [ -d $MOUNT ] || mkdir -p $MOUNT grep " $MOUNT " /etc/mtab && echo "test 5c: mtab before mount" && return 10 local oldfs="${FSNAME}" @@ -335,8 +357,8 @@ test_5d() { run_test 5d "mount with ost down" test_5e() { - start_ost start_mds + start_ost #define OBD_FAIL_PTLRPC_DELAY_SEND 0x506 do_facet client "lctl set_param fail_loc=0x80000506" @@ -382,7 +404,7 @@ test_9() { CHECK_PTLDEBUG="`do_facet ost1 lctl get_param -n debug`" if [ "$CHECK_PTLDEBUG" ] && { \ [ "$CHECK_PTLDEBUG" = "trace inode warning error emerg console" ] || - [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then + [ "$CHECK_PTLDEBUG" = "trace inode" ]; }; then echo "lnet.debug success" else echo "lnet.debug: want 'trace inode', have '$CHECK_PTLDEBUG'" @@ -403,8 +425,8 @@ run_test 9 "test ptldebug and subsystem for mkfs" # LOGS/PENDING do not exist anymore since CMD3 test_16() { local TMPMTPT="${TMP}/conf16" - local dev=${SINGLEMDS}_dev - local MDSDEV=${!dev} + local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) + if [ ! -e "$MDSDEV" ]; then log "no $MDSDEV existing, so mount Lustre to create one" setup @@ -437,8 +459,7 @@ test_16() { run_test 16 "verify that lustre will correct the mode of OBJECTS" test_17() { - local dev=${SINGLEMDS}_dev - local MDSDEV=${!dev} + local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) if [ ! -e "$MDSDEV" ]; then echo "no $MDSDEV existing, so mount Lustre to create one" @@ -452,15 +473,14 @@ test_17() { start_ost start_mds && return 42 - gen_config + reformat_and_config } run_test 17 "Verify failed mds_postsetup won't fail assertion (2936) (should return errs)" test_18() { [ "$FSTYPE" != "ldiskfs" ] && skip "not needed for FSTYPE=$FSTYPE" && return - local dev=${SINGLEMDS}_dev - local MDSDEV=${!dev} + local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) local MIN=2000000 @@ -487,7 +507,7 @@ test_18() { [ $SPACE -gt $((MIN / 20)) ] && OK=1 && myMDSSIZE=$MIN && \ log "use file $MDSDEV with MIN=$MIN" - [ -z "$OK" ] && skip "$MDSDEV too small for ${MIN}kB MDS" && return + [ -z "$OK" ] && skip_env "$MDSDEV too small for ${MIN}kB MDS" && return echo "mount mds with large journal..." @@ -495,7 +515,7 @@ test_18() { MDS_MKFS_OPTS="--mgs --mdt --fsname=$FSNAME --device-size=$myMDSSIZE --param sys.timeout=$TIMEOUT $MDSOPT" - gen_config + reformat_and_config echo "mount lustre system..." setup check_mount || return 41 @@ -511,7 +531,7 @@ test_18() { cleanup || return $? MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS - gen_config + reformat_and_config } run_test 18 "check mkfs creates large journals" @@ -529,8 +549,8 @@ run_test 19b "start/stop OSTs without MDS" test_20() { # first format the ost/mdt - start_ost start_mds + start_ost mount_client $MOUNT check_mount || return 43 rm -f $DIR/$tfile @@ -551,6 +571,7 @@ run_test 20 "remount ro,rw mounts work and doesn't break /etc/mtab" test_21a() { start_mds start_ost + wait_osc_import_state mds ost FULL stop_ost stop_mds } @@ -559,6 +580,7 @@ run_test 21a "start mds before ost, stop ost first" test_21b() { start_ost start_mds + wait_osc_import_state mds ost FULL stop_mds stop_ost } @@ -568,6 +590,7 @@ test_21c() { start_ost start_mds start_ost2 + wait_osc_import_state mds ost2 FULL stop_ost stop_ost2 stop_mds @@ -582,11 +605,12 @@ test_22() { echo Client mount with ost in logs, but none running start_ost # wait until mds connected to ost and open client connection - # ping_interval + 1 - sleep $((TIMEOUT / 4 + 1)) + wait_osc_import_state mds ost FULL stop_ost mount_client $MOUNT # check_mount will block trying to contact ost + mcreate $DIR/$tfile || return 40 + rm -f $DIR/$tfile || return 42 umount_client $MOUNT pass @@ -609,9 +633,9 @@ test_22() { run_test 22 "start a client before osts (should return errs)" test_23a() { # was test_23 - setup - # fail mds - stop $SINGLEMDS + setup + # fail mds + stop $SINGLEMDS # force down client so that recovering mds waits for reconnect local running=$(grep -c $MOUNT /proc/mounts) || true if [ $running -ne 0 ]; then @@ -637,7 +661,7 @@ test_23a() { # was test_23 local PID1 local PID2 local WAIT=0 - local MAX_WAIT=20 + local MAX_WAIT=30 local sleep=1 while [ "$WAIT" -lt "$MAX_WAIT" ]; do sleep $sleep @@ -649,9 +673,11 @@ test_23a() { # was test_23 echo "waiting for mount to finish ... " WAIT=$(( WAIT + sleep)) done - [ "$WAIT" -eq "$MAX_WAIT" ] && error "MOUNT_PID $MOUNT_PID and \ - MOUNT__LUSTRE_PID $MOUNT__LUSTRE_PID still not killed in $WAIT secs" - ps -ef | grep mount + if [ "$WAIT" -eq "$MAX_WAIT" ]; then + error "MOUNT_PID $MOUNT_PID and "\ + "MOUNT_LUSTRE_PID $MOUNT_LUSTRE_PID still not killed in $WAIT secs" + ps -ef | grep mount + fi stop_mds || error stop_ost || error } @@ -661,8 +687,8 @@ umount_client $MOUNT cleanup_nocli test_23b() { # was test_23 - start_ost start_mds + start_ost # Simulate -EINTR during mount OBD_FAIL_LDLM_CLOSE_THREAD lctl set_param fail_loc=0x80000313 mount_client $MOUNT @@ -684,13 +710,16 @@ cleanup_24a() { } test_24a() { - #set up fs1 + #set up fs1 gen_config + #set up fs2 + local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) + [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \ - skip "mixed loopback and real device not working" && return + skip_env "mixed loopback and real device not working" && return fi local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} @@ -719,11 +748,11 @@ test_24a() { rm $MOUNT2/b || return 4 # 2 is actually mounted grep $MOUNT2' ' /proc/mounts > /dev/null || return 5 - # failover + # failover 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 $4 }' | head -1) @@ -734,14 +763,16 @@ test_24a() { run_test 24a "Multiple MDTs on a single node" test_24b() { + local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) + if [ -z "$fs2mds_DEV" ]; then do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \ - skip "mixed loopback and real device not working" && return + skip_env "mixed loopback and real device not working" && return fi local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} - add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10 + add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME}2 --mgs --reformat $fs2mdsdev || exit 10 setup start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && return 2 cleanup || return 6 @@ -775,7 +806,7 @@ set_and_check() { local myfacet=$1 local TEST=$2 local PARAM=$3 - local ORIG=$(do_facet $myfacet "$TEST") + local ORIG=$(do_facet $myfacet "$TEST") if [ $# -gt 3 ]; then local FINAL=$4 else @@ -783,16 +814,16 @@ set_and_check() { FINAL=$(($ORIG + 5)) fi echo "Setting $PARAM from $ORIG to $FINAL" - do_facet $SINGLEMDS "$LCTL conf_param $PARAM=$FINAL" || error conf_param failed + do_facet $SINGLEMDS "$LCTL conf_param $PARAM='$FINAL'" || error conf_param failed - wait_update $(facet_host $myfacet) "$TEST" $FINAL || error check failed! + wait_update $(facet_host $myfacet) "$TEST" "$FINAL" || error check failed! } test_27a() { start_ost || return 1 start_mds || return 2 - echo "Requeue thread should have started: " - ps -e | grep ll_cfg_requeue + echo "Requeue thread should have started: " + ps -e | grep ll_cfg_requeue set_and_check ost1 "lctl get_param -n obdfilter.$FSNAME-OST0000.client_cache_seconds" "$FSNAME-OST0000.ost.client_cache_seconds" || return 3 cleanup_nocli } @@ -835,7 +866,7 @@ test_28() { run_test 28 "permanent parameter setting" test_29() { - [ "$OSTCOUNT" -lt "2" ] && skip "$OSTCOUNT < 2, skipping" && return + [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return setup > /dev/null 2>&1 start_ost2 sleep 10 @@ -856,7 +887,7 @@ test_29() { echo "Live client success: got $RESULT" fi - # check MDT too + # check MDT too local MPROC="osc.$FSNAME-OST0001-osc-[M]*.active" local MAX=30 local WAIT=0 @@ -895,30 +926,70 @@ test_29() { cleanup_nocli #writeconf to remove all ost2 traces for subsequent tests writeconf - start_mds - start_ost - sleep 5 - cleanup } run_test 29 "permanently remove an OST" -test_30() { +test_30a() { setup + echo Big config llog TEST="lctl get_param -n llite.$FSNAME-*.max_read_ahead_whole_mb" ORIG=$($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 done - # make sure client restart still works + # make sure client restart still works umount_client $MOUNT mount_client $MOUNT || return 4 - [ "$($TEST)" -ne "$i" ] && return 5 - set_and_check client "$TEST" "$FSNAME.llite.max_read_ahead_whole_mb" $ORIG || return 6 + [ "$($TEST)" -ne "$i" ] && error "Param didn't stick across restart $($TEST) != $i" + pass + + echo Erase parameter setting + do_facet mgs "$LCTL conf_param -d $FSNAME.llite.max_read_ahead_whole_mb" || return 6 + umount_client $MOUNT + mount_client $MOUNT || return 6 + FINAL=$($TEST) + echo "deleted (default) value=$FINAL, orig=$ORIG" + # assumes this parameter started at the default value + [ "$FINAL" -eq "$ORIG" ] || fail "Deleted value=$FINAL, orig=$ORIG" + cleanup } -run_test 30 "Big config llog" +run_test 30a "Big config llog and conf_param deletion" + +test_30b() { + setup + + # Make a fake nid. Use the OST nid, and add 20 to the least significant + # numerical part of it. Hopefully that's not already a failover address for + # the server. + OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | awk '{print $1}') + ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@") + NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256)) + NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/") + 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 $NEW" + NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids) + echo $NIDS + NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1)) + echo "should have 2 failover nids: $NIDCOUNT" + [ $NIDCOUNT -eq 2 ] || error "Failover nid not added" + do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" || error "conf_param delete failed" + umount_client $MOUNT + mount_client $MOUNT || return 3 + + NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids) + echo $NIDS + NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 1)) + echo "only 1 final nid should remain: $NIDCOUNT" + [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed" + + cleanup +} +run_test 30b "Remove failover nids" test_31() { # bug 10734 # ipaddr must not exist @@ -928,7 +999,7 @@ test_31() { # bug 10734 run_test 31 "Connect to non-existent node (shouldn't crash)" # Use these start32/stop32 fn instead of t-f start/stop fn, -# for local devices, to skip global facet vars init +# for local devices, to skip global facet vars init stop32 () { local facet=$1 shift @@ -946,11 +1017,11 @@ start32 () { echo "Starting local ${facet}: $@ $device ${MOUNT%/*}/${facet}" mount -t lustre $@ ${device} ${MOUNT%/*}/${facet} - RC=$? + local RC=$? if [ $RC -ne 0 ]; then echo "mount -t lustre $@ ${device} ${MOUNT%/*}/${facet}" echo "Start of ${device} of local ${facet} failed ${RC}" - fi + fi losetup -a return $RC } @@ -971,66 +1042,59 @@ cleanup_32() { } test_32a() { - # this test is totally useless on a client-only system - [ -n "$CLIENTONLY" -o -n "$CLIENTMODSONLY" ] && skip "client only testing" && return 0 + client_only && skip "client only testing" && return 0 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; } - [ -z "$TUNEFS" ] && skip "No tunefs" && return 0 - local DISK1_8=$LUSTRE/tests/disk1_8.tgz - [ ! -r $DISK1_8 ] && skip "Cannot find $DISK1_8" && return 0 + [ -z "$TUNEFS" ] && skip_env "No tunefs" && return 0 - mkdir -p $TMP/$tdir - tar xjvf $DISK1_8 -C $TMP/$tdir || \ - { skip "Cannot untar $DISK1_8" && return 0; } + local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2 + [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0 + local tmpdir=$TMP/conf32a + mkdir -p $tmpdir + tar xjvf $DISK1_8 -C $tmpdir || \ + { 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" # nids are wrong, so client wont work, but server should start - start32 mds $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \ + start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \ trap cleanup_32 EXIT INT || return 3 - - local UUID=$(lctl get_param -n mds.lustre-MDT0000.uuid) + + local UUID=$($LCTL get_param -n mdt.lustre-MDT0000.uuid) echo MDS uuid $UUID - [ "$UUID" == "mdsA_UUID" ] || error "UUID is wrong: $UUID" + [ "$UUID" == "lustre-MDT0000_UUID" ] || error "UUID is wrong: $UUID" - $TUNEFS --mgsnode=`hostname` $tmpdir/ost1 || error "tunefs failed" + $TUNEFS --mgsnode=$HOSTNAME $tmpdir/ost1 || error "tunefs failed" start32 ost1 $tmpdir/ost1 "-o loop" || return 5 - UUID=$(lctl get_param -n obdfilter.lustre-OST0000.uuid) + UUID=$($LCTL get_param -n obdfilter.lustre-OST0000.uuid) echo OST uuid $UUID - [ "$UUID" == "ost1_UUID" ] || error "UUID is wrong: $UUID" + [ "$UUID" == "lustre-OST0000_UUID" ] || error "UUID is wrong: $UUID" local NID=$($LCTL list_nids | head -1) - echo "OSC changes should return err:" - $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 && return 7 - $LCTL conf_param lustre-OST0000.failover.node=$NID && return 8 + echo "OSC changes should succeed:" + $LCTL conf_param lustre-OST0000.osc.max_dirty_mb=15 || return 7 + $LCTL conf_param lustre-OST0000.failover.node=$NID || return 8 echo "ok." - echo "MDC changes should succeed:" + + echo "MDC changes should succeed:" $LCTL conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=9 || return 9 $LCTL conf_param lustre-MDT0000.failover.node=$NID || return 10 echo "ok." - # With a new good MDT failover nid, we should be able to mount a client - # (but it cant talk to OST) - local mountopt="-o exclude=lustre-OST0000" - - local device=`h2$NETTYPE $HOSTNAME`:/lustre - echo "Starting local client: $HOSTNAME: $mountopt $device $MOUNT" - mount -t lustre $mountopt $device $MOUNT || return 1 - - local old=$(lctl get_param -n mdc.*.max_rpcs_in_flight) - local new=$((old + 5)) - lctl conf_param lustre-MDT0000.mdc.max_rpcs_in_flight=$new - wait_update $HOSTNAME "lctl get_param -n mdc.*.max_rpcs_in_flight" $new || return 11 + echo "LOV changes should succeed:" + $LCTL pool_new lustre.interop || return 11 + $LCTL conf_param lustre-MDT0000.lov.stripesize=4M || return 12 + echo "ok." cleanup_32 # mount a second time to make sure we didnt leave upgrade flag on load_modules $TUNEFS --dryrun $tmpdir/mds || error "tunefs failed" - start32 mds $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \ + start32 mds1 $tmpdir/mds "-o loop,exclude=lustre-OST0000" && \ trap cleanup_32 EXIT INT || return 12 cleanup_32 @@ -1040,20 +1104,19 @@ test_32a() { run_test 32a "Upgrade from 1.8 (not live)" test_32b() { - # this test is totally useless on a client-only system - [ -n "$CLIENTONLY" -o -n "$CLIENTMODSONLY" ] && skip "client only testing" && return 0 + client_only && skip "client only testing" && return 0 [ "$NETTYPE" = "tcp" ] || { skip "NETTYPE != tcp" && return 0; } - [ -z "$TUNEFS" ] && skip "No tunefs" && return + [ -z "$TUNEFS" ] && skip_env "No tunefs" && return - local DISK1_8=$LUSTRE/tests/disk1_8.tgz - [ ! -r $DISK1_8 ] && skip "Cannot find $DISK1_8" && return 0 - local tmpdir=$TMP/$tdir + local DISK1_8=$LUSTRE/tests/disk1_8.tar.bz2 + [ ! -r $DISK1_8 ] && skip_env "Cannot find $DISK1_8" && return 0 + local tmpdir=$TMP/conf32b mkdir -p $tmpdir tar xjvf $DISK1_8 -C $tmpdir || \ - { skip "Cannot untar $DISK1_8" && return ; } + { skip_env "Cannot untar $DISK1_8" && return ; } load_modules - lctl set_param debug=$PTLDEBUG + $LCTL set_param debug="config" local NEWNAME=lustre # writeconf will cause servers to register with their current nids @@ -1061,22 +1124,32 @@ test_32b() { start32 mds1 $tmpdir/mds "-o loop" && \ trap cleanup_32 EXIT INT || return 3 - local UUID=$(lctl get_param -n mdt.${NEWNAME}-MDT0000.uuid) + local UUID=$($LCTL get_param -n mdt.${NEWNAME}-MDT0000.uuid) echo MDS uuid $UUID - [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID" + [ "$UUID" == "${NEWNAME}-MDT0000_UUID" ] || error "UUID is wrong: $UUID" - $TUNEFS --mgsnode=`hostname` --writeconf --fsname=$NEWNAME $tmpdir/ost1 || error "tunefs failed" + $TUNEFS --mgsnode=$HOSTNAME --writeconf --fsname=$NEWNAME $tmpdir/ost1 ||\ + error "tunefs failed" start32 ost1 $tmpdir/ost1 "-o loop" || return 5 - UUID=$(lctl get_param -n obdfilter.${NEWNAME}-OST0000.uuid) + UUID=$($LCTL get_param -n obdfilter.${NEWNAME}-OST0000.uuid) echo OST uuid $UUID [ "$UUID" == "${NEWNAME}-OST0000_UUID" ] || error "UUID is wrong: $UUID" - echo "OSC changes should succeed:" + local NID=$($LCTL list_nids | head -1) + + echo "OSC changes should succeed:" $LCTL conf_param ${NEWNAME}-OST0000.osc.max_dirty_mb=15 || return 7 $LCTL conf_param ${NEWNAME}-OST0000.failover.node=$NID || return 8 echo "ok." - echo "MDC changes should succeed:" + + echo "MDC changes should succeed:" $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=9 || return 9 + $LCTL conf_param ${NEWNAME}-MDT0000.failover.node=$NID || return 10 + echo "ok." + + echo "LOV changes should succeed:" + $LCTL pool_new ${NEWNAME}.interop || return 11 + $LCTL conf_param ${NEWNAME}-MDT0000.lov.stripesize=4M || return 12 echo "ok." # MDT and OST should have registered with new nids, so we should have @@ -1087,10 +1160,10 @@ test_32b() { echo "Starting local client: $HOSTNAME: $device $MOUNT" mount -t lustre $device $MOUNT || return 1 - local old=$(lctl get_param -n mdc.*.max_rpcs_in_flight) + local old=$($LCTL get_param -n mdc.*.max_rpcs_in_flight) local new=$((old + 5)) - lctl conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new - wait_update $HOSTNAME "lctl get_param -n mdc.*.max_rpcs_in_flight" $new || return 11 + $LCTL conf_param ${NEWNAME}-MDT0000.mdc.max_rpcs_in_flight=$new + wait_update $HOSTNAME "$LCTL get_param -n mdc.*.max_rpcs_in_flight" $new || return 11 [ "$(cksum $MOUNT/passwd | cut -d' ' -f 1,2)" == "94306271 1478" ] || return 12 echo "ok." @@ -1104,16 +1177,18 @@ run_test 32b "Upgrade from 1.8 with writeconf" test_33a() { # bug 12333, was test_33 local rc=0 local FSNAME2=test-123 + local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) + [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" ]; then do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \ - skip "mixed loopback and real device not working" && return + skip_env "mixed loopback and real device not working" && return fi local fs2mdsdev=${fs2mds_DEV:-${MDSDEV}_2} local fs2ostdev=${fs2ost_DEV:-$(ostdevname 1)_2} - add fs2mds $MDS_MKFS_OPTS --fsname=${FSNAME2} --reformat $fs2mdsdev || exit 10 + 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 start fs2mds $fs2mdsdev $MDS_MOUNT_OPTS && trap cleanup_24a EXIT INT @@ -1123,7 +1198,7 @@ test_33a() { # bug 12333, was test_33 mount -t lustre $MGSNID:/${FSNAME2} $MOUNT2 || rc=2 echo "ok." - cp /etc/hosts $MOUNT2/ || rc=3 + cp /etc/hosts $MOUNT2/ || rc=3 $LFS getstripe $MOUNT2/hosts umount -d $MOUNT2 @@ -1195,7 +1270,7 @@ test_34c() { } run_test 34c "force umount with failed ost should be normal" -test_35() { # bug 12459 +test_35a() { # bug 12459 setup DBG_SAVE="`lctl get_param -n debug`" @@ -1209,7 +1284,7 @@ test_35() { # bug 12459 log "Wait for RECONNECT_INTERVAL seconds (10s)" sleep 10 - MSG="conf-sanity.sh test_35 `date +%F%kh%Mm%Ss`" + MSG="conf-sanity.sh test_35a `date +%F%kh%Mm%Ss`" $LCTL clear log "$MSG" log "Stopping the MDT:" @@ -1240,21 +1315,89 @@ test_35() { # bug 12459 [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7 cleanup } -run_test 35 "Reconnect to the last active server first" +run_test 35a "Reconnect to the last active server first" + +test_35b() { # bug 18674 + remote_mds || { skip "local MDS" && return 0; } + setup + + debugsave + $LCTL set_param debug="ha" + $LCTL clear + MSG="conf-sanity.sh test_35b `date +%F%kh%Mm%Ss`" + log "$MSG" + + log "Set up a fake failnode for the MDS" + FAKENID="127.0.0.2" + local device=$(do_facet mds "$LCTL get_param -n devices" | \ + awk '($3 ~ "mdt" && $4 ~ "MDT") { print $4 }' | head -1) + do_facet mds "$LCTL conf_param ${device}.failover.node=$FAKENID" || \ + return 1 + + local at_max_saved=0 + # 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 $MOUNT/testdir/test + + log "Injecting EBUSY on MDS" + # Setting OBD_FAIL_MDS_RESEND=0x136 + do_facet mds "$LCTL set_param fail_loc=0x80000136" || return 2 + + log "Stat on a test file" + stat $MOUNT/testdir/test + + log "Stop injecting EBUSY on MDS" + do_facet mds "$LCTL set_param fail_loc=0" || return 3 + rm -f $MOUNT/testdir/test + + log "done" + # restore adaptive timeout + [ $at_max_saved -ne 0 ] && at_max_set $at_max_saved mds client + + $LCTL dk $TMP/lustre-log-$TESTNAME.log + + # 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;} + /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ { + ret = 1; + if (\\\$NF ~ /$FAKENID/) { + ret = 2; + exit; + } + } + END {print ret}" $TMP/lustre-log-$TESTNAME.log` + + [ "$FAILCONN" == "0" ] && \ + log "ERROR: The client reconnection has not been triggered" && \ + return 4 + [ "$FAILCONN" == "2" ] && \ + log "ERROR: The client tried to reconnect to the failover server while the primary was busy" && \ + return 5 + + cleanup +} +run_test 35b "Continue reconnection retries, if the active server is busy" test_36() { # 12743 local rc local FSNAME2=test1234 local fs3ost_HOST=$ost_HOST + local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) [ -n "$ost1_HOST" ] && fs2ost_HOST=$ost1_HOST && fs3ost_HOST=$ost1_HOST rc=0 if [ -z "$fs2ost_DEV" -o -z "$fs2mds_DEV" -o -z "$fs3ost_DEV" ]; then do_facet $SINGLEMDS [ -b "$MDSDEV" ] && \ - skip "mixed loopback and real device not working" && return + skip_env "mixed loopback and real device not working" && return fi - [ $OSTCOUNT -lt 2 ] && skip "skipping test for single OST" && return + [ $OSTCOUNT -lt 2 ] && skip_env "skipping test for single OST" && return [ "$ost_HOST" = "`hostname`" -o "$ost1_HOST" = "`hostname`" ] || \ { skip "remote OST" && return 0; } @@ -1289,17 +1432,17 @@ test_36() { # 12743 ALLOWANCE=$((64 * $OSTCOUNT)) - if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] || + if [ $DFTOTAL -lt $(($BKTOTAL - $ALLOWANCE)) ] || [ $DFTOTAL -gt $(($BKTOTAL + $ALLOWANCE)) ] ; then echo "**** FAIL: df total($DFTOTAL) mismatch OST total($BKTOTAL)" rc=1 fi - if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] || + if [ $DFFREE -lt $(($BKFREE - $ALLOWANCE)) ] || [ $DFFREE -gt $(($BKFREE + $ALLOWANCE)) ] ; then echo "**** FAIL: df free($DFFREE) mismatch OST free($BKFREE)" rc=2 fi - if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] || + if [ $DFAVAIL -lt $(($BKAVAIL - $ALLOWANCE)) ] || [ $DFAVAIL -gt $(($BKAVAIL + $ALLOWANCE)) ] ; then echo "**** FAIL: df avail($DFAVAIL) mismatch OST avail($BKAVAIL)" rc=3 @@ -1316,7 +1459,7 @@ test_36() { # 12743 run_test 36 "df report consistency on OSTs with different block size" test_37() { - [ -n "$CLIENTONLY" -o -n "$CLIENTMODSONLY" ] && skip "client only testing" && return 0 + client_only && skip "client only testing" && return 0 LOCAL_MDSDEV="$TMP/mdt.img" SYM_MDSDEV="$TMP/sym_mdt.img" @@ -1358,8 +1501,7 @@ test_38() { # bug 14222 log "rename lov_objid file on MDS" rm -f $TMP/lov_objid.orig - local dev=${SINGLEMDS}_dev - local MDSDEV=${!dev} + local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) do_facet $SINGLEMDS "$DEBUGFS -c -R \\\"dump lov_objid $TMP/lov_objid.orig\\\" $MDSDEV" do_facet $SINGLEMDS "$DEBUGFS -w -R \\\"rm lov_objid\\\" $MDSDEV" @@ -1404,7 +1546,7 @@ test_39() { PTLDEBUG=+malloc setup cleanup - perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' && + perl $SRCDIR/leak_finder.pl $TMP/debug 2>&1 | egrep '*** Leak:' && error "memory leak detected" || true } run_test 39 "leak_finder recognizes both LUSTRE and LNET malloc messages" @@ -1420,12 +1562,11 @@ run_test 40 "race during service thread startup" test_41() { #bug 14134 local rc - local dev=${SINGLEMDS}_dev - local MDSDEV=${!dev} + local MDSDEV=$(mdsdevname ${SINGLEMDS//mds/}) start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nosvc -n start ost1 `ostdevname 1` $OST_MOUNT_OPTS - start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs + start $SINGLEMDS $MDSDEV $MDS_MOUNT_OPTS -o nomgs,force mkdir -p $MOUNT mount_client $MOUNT || return 1 sleep 5 @@ -1445,7 +1586,7 @@ run_test 41 "mount mds with --nosvc and --nomgs" test_42() { #bug 14693 setup check_mount || return 2 - do_facet client lctl conf_param lustre.llite.some_wrong_param=10 + do_facet mgs $LCTL conf_param lustre.llite.some_wrong_param=10 umount_client $MOUNT mount_client $MOUNT || return 1 cleanup @@ -1453,9 +1594,127 @@ 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" + + # + # create set of test files + # + echo "111" > $DIR/$tfile-userfile || error "write 1 failed" + chmod go-rw $DIR/$tfile-userfile || error "chmod 1 failed" + chown $RUNAS_ID.$RUNAS_ID $DIR/$tfile-userfile || error "chown failed" + + echo "222" > $DIR/$tfile-rootfile || error "write 2 failed" + chmod go-rw $DIR/$tfile-rootfile || error "chmod 2 faield" + + mkdir $DIR/$tdir-rootdir -p || error "mkdir failed" + 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" + + ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-userfile) + dd if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \ + error "$ST: root read permission is denied" + echo "$ST: root read permission is granted - ok" + + echo "444" | \ + dd conv=notrunc if=$DIR/$tfile-userfile 1>/dev/null 2>/dev/null || \ + error "$ST: root write permission is denied" + echo "$ST: root write permission is granted - ok" + + ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile) + dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \ + error "$ST: root read permission is granted" + echo "$ST: root read permission is denied - ok" + + echo "555" | \ + dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null && \ + error "$ST: root write permission is granted" + echo "$ST: root write permission is denied - ok" + + ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir) + rm $DIR/$tdir-rootdir/tfile-1 1>/dev/null 2>/dev/null && \ + error "$ST: root unlink permission is granted" + echo "$ST: root unlink permission is denied - ok" + + touch $DIR/tdir-rootdir/tfile-2 1>/dev/null 2>/dev/null && \ + error "$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" + + ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tfile-rootfile) + dd if=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \ + error "$ST: root read permission is denied" + echo "$ST: root read permission is granted - ok" + + echo "666" | \ + dd conv=notrunc of=$DIR/$tfile-rootfile 1>/dev/null 2>/dev/null || \ + error "$ST: root write permission is denied" + echo "$ST: root write permission is granted - ok" + + ST=$(stat -c "%n: owner uid %u (%A)" $DIR/$tdir-rootdir) + rm $DIR/$tdir-rootdir/tfile-1 || \ + error "$ST: root unlink permission is denied" + echo "$ST: root unlink permission is granted - ok" + touch $DIR/$tdir-rootdir/tfile-2 || \ + error "$ST: root create permission is denied" + echo "$ST: root create permission is granted - ok" + + return 0 +} +run_test 43 "check root_squash and nosquash_nids" + umount_client $MOUNT cleanup_nocli +test_44() { # 16317 + setup + check_mount || return 2 + UUID=$($LCTL get_param llite.${FSNAME}*.uuid | cut -d= -f2) + STATS_FOUND=no + UUIDS=$(do_facet mds "$LCTL get_param mdt.${FSNAME}*.exports.*.uuid") + for VAL in $UUIDS; do + NID=$(echo $VAL | cut -d= -f1) + CLUUID=$(echo $VAL | cut -d= -f2) + [ "$UUID" = "$CLUUID" ] && STATS_FOUND=yes && break + done + [ "$STATS_FOUND" = "no" ] && error "stats not found for client" + cleanup + return 0 +} +run_test 44 "mounted client proc entry exists" + test_45() { #17310 setup check_mount || return 2 @@ -1476,23 +1735,48 @@ test_45() { #17310 } run_test 45 "long unlink handling in ptlrpcd" +cleanup_46a() { + trap 0 + local rc=0 + local count=$1 + + umount_client $MOUNT2 || rc=$? + umount_client $MOUNT || rc=$? + while [ $count -gt 0 ]; do + stop ost${count} -f || rc=$? + let count=count-1 + done + stop_mds || rc=$? + # writeconf is needed after the test, otherwise, + # we might end up with extra OSTs + writeconf || rc=$? + cleanup_nocli || rc=$? + return $rc +} + test_46a() { - OSTCOUNT=6 - reformat + echo "Testing with $OSTCOUNT OSTs" + reformat_and_config start_mds || return 1 #first client should see only one ost start_ost || return 2 + wait_osc_import_state mds ost FULL #start_client mount_client $MOUNT || return 3 - - start_ost2 || return 4 - start ost3 `ostdevname 3` $OST_MOUNT_OPTS || return 5 - start ost4 `ostdevname 4` $OST_MOUNT_OPTS || return 6 - start ost5 `ostdevname 5` $OST_MOUNT_OPTS || return 7 - # wait until ost2-5 is sync - # ping_interval + 1 - sleep $((TIMEOUT / 4 + 1)) - #second client see both ost's + trap "cleanup_46a $OSTCOUNT" EXIT ERR + + local i + for (( i=2; i<=$OSTCOUNT; i++ )); do + start ost$i `ostdevname $i` $OST_MOUNT_OPTS || return $((i+2)) + done + + # wait until osts in sync + for (( i=2; i<=$OSTCOUNT; i++ )); do + wait_osc_import_state mds ost$i FULL + done + + + #second client see all ost's mount_client $MOUNT2 || return 8 $LFS setstripe $MOUNT2 -c -1 || return 9 @@ -1501,21 +1785,15 @@ test_46a() { echo "ok" > $MOUNT2/widestripe $LFS getstripe $MOUNT2/widestripe || return 11 # fill acl buffer for avoid expand lsm to them - awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do + awk -F : '{if (FNR < 25) { print "u:"$1":rwx" }}' /etc/passwd | while read acl; do setfacl -m $acl $MOUNT2/widestripe done # will be deadlock stat $MOUNT/widestripe || return 12 - umount_client $MOUNT2 || return 13 - umount_client $MOUNT || return 14 - stop ost5 -f || return 20 - stop ost4 -f || return 21 - stop ost3 -f || return 22 - stop_ost2 || return 23 - stop_ost || return 24 - stop_mds || return 25 + cleanup_46a $OSTCOUNT || { echo "cleanup_46a failed!" && return 13; } + return 0 } run_test 46a "handle ost additional - wide striped file" @@ -1535,10 +1813,10 @@ test_47() { #17674 lru_size[count]=$lrs let count=count+1 done - + facet_failover ost1 facet_failover $SINGLEMDS - df -h $MOUNT || return 3 + client_up || return 3 count=0 for ns in $($LCTL get_param ldlm.namespaces.$FSNAME-*-*-*.lru_size); do @@ -1558,8 +1836,14 @@ test_47() { #17674 } run_test 47 "server restart does not make client loss lru_resize settings" -# reformat after this test must need - if test will failed -# we will have unkillable file at FS +cleanup_48() { + trap 0 + + # reformat after this test is needed - if test will failed + # we will have unkillable file at FS + reformat_and_config +} + test_48() { # bug 17636 reformat setup_noconfig @@ -1570,22 +1854,606 @@ test_48() { # bug 17636 echo "ok" > $MOUNT/widestripe $LFS getstripe $MOUNT/widestripe || return 11 + + trap cleanup_48 EXIT ERR + # fill acl buffer for avoid expand lsm to them - awk -F : '{ print "u:"$1":rwx" }' /etc/passwd | while read acl; do + getent passwd | awk -F : '{ print "u:"$1":rwx" }' | while read acl; do setfacl -m $acl $MOUNT/widestripe done - awk -F : '{ print "g:"$1":rwx" }' /etc/groups | while read acl; do - setfacl -m $acl $MOUNT/widestripe - done - stat $MOUNT/widestripe || return 12 - - cleanup || error "can't cleanup" + + cleanup_48 return 0 } run_test 48 "too many acls on file" +# check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE +test_49() { # bug 17710 + local OLD_MDS_MKFS_OPTS=$MDS_MKFS_OPTS + local OLD_OST_MKFS_OPTS=$OST_MKFS_OPTS + local LOCAL_TIMEOUT=20 + + + OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$LOCAL_TIMEOUT $MKFSOPT $OSTOPT" + + reformat + setup_noconfig + check_mount || return 1 + + echo "check ldlm_timout..." + LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`" + LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`" + LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`" + + if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then + error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT" + fi + + if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then + error "LDLM_TIMEOUT($LDLM_MDS) is not correct" + fi + + umount_client $MOUNT + stop_ost || return 2 + stop_mds || return 3 + + OST_MKFS_OPTS="--ost --fsname=$FSNAME --device-size=$OSTSIZE --mgsnode=$MGSNID --param sys.timeout=$LOCAL_TIMEOUT --param sys.ldlm_timeout=$((LOCAL_TIMEOUT - 1)) $MKFSOPT $OSTOPT" + + reformat + setup_noconfig + check_mount || return 7 + + LDLM_MDS="`do_facet mds lctl get_param -n ldlm_timeout`" + LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`" + LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`" + + if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then + error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT" + fi + + if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then + error "LDLM_TIMEOUT($LDLM_MDS) is not correct" + fi + + cleanup || return $? + + MDS_MKFS_OPTS=$OLD_MDS_MKFS_OPTS + OST_MKFS_OPTS=$OLD_OST_MKFS_OPTS +} +run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE" + +lazystatfs() { + # Test both statfs and lfs df and fail if either one fails + multiop_bg_pause $1 f_ + RC1=$? + PID=$! + killall -USR1 multiop + [ $RC1 -ne 0 ] && log "lazystatfs multiop failed" + wait $PID || { RC1=$?; log "multiop return error "; } + + $LFS df & + PID=$! + sleep 5 + kill -s 0 $PID + RC2=$? + if [ $RC2 -eq 0 ]; then + kill -s 9 $PID + log "lazystatfs df failed" + fi + + RC=0 + [[ $RC1 -ne 0 || $RC2 -eq 0 ]] && RC=1 + return $RC +} + +test_50a() { + setup + lctl set_param llite.$FSNAME-*.lazystatfs=1 + touch $DIR/$tfile + + lazystatfs $MOUNT || error "lazystatfs failed but no down servers" + + cleanup || return $? +} +run_test 50a "lazystatfs all servers available ==========================" + +test_50b() { + setup + lctl set_param llite.$FSNAME-*.lazystatfs=1 + touch $DIR/$tfile + + # Wait for client to detect down OST + stop_ost || error "Unable to stop OST1" + wait_osc_import_state mds ost DISCONN + + lazystatfs $MOUNT || error "lazystatfs should don't have returned EIO" + + umount_client $MOUNT || error "Unable to unmount client" + stop_mds || error "Unable to stop MDS" +} +run_test 50b "lazystatfs all servers down ==========================" + +test_50c() { + start_mds || error "Unable to start MDS" + start_ost || error "Unable to start OST1" + start_ost2 || error "Unable to start OST2" + mount_client $MOUNT || error "Unable to mount client" + lctl set_param llite.$FSNAME-*.lazystatfs=1 + touch $DIR/$tfile + + # Wait for client to detect down OST + stop_ost || error "Unable to stop OST1" + wait_osc_import_state mds ost DISCONN + lazystatfs $MOUNT || error "lazystatfs failed with one down server" + + umount_client $MOUNT || error "Unable to unmount client" + stop_ost2 || error "Unable to stop OST2" + stop_mds || error "Unable to stop MDS" +} +run_test 50c "lazystatfs one server down ==========================" + +test_50d() { + start_mds || error "Unable to start MDS" + start_ost || error "Unable to start OST1" + start_ost2 || error "Unable to start OST2" + mount_client $MOUNT || error "Unable to mount client" + lctl set_param llite.$FSNAME-*.lazystatfs=1 + touch $DIR/$tfile + + # Issue the statfs during the window where the client still + # belives the OST to be available but it is in fact down. + # No failure just a statfs which hangs for a timeout interval. + stop_ost || error "Unable to stop OST1" + lazystatfs $MOUNT || error "lazystatfs failed with one down server" + + umount_client $MOUNT || error "Unable to unmount client" + stop_ost2 || error "Unable to stop OST2" + stop_mds || error "Unable to stop MDS" +} +run_test 50d "lazystatfs client/server conn race ==========================" + +test_50e() { + local RC1 + local pid + + reformat_and_config + start_mds || return 1 + #first client should see only one ost + start_ost || return 2 + wait_osc_import_state mds ost FULL + + # Wait for client to detect down OST + stop_ost || error "Unable to stop OST1" + wait_osc_import_state mds ost DISCONN + + mount_client $MOUNT || error "Unable to mount client" + lctl set_param llite.$FSNAME-*.lazystatfs=0 + + multiop_bg_pause $MOUNT _f + RC1=$? + pid=$! + + if [ $RC1 -ne 0 ]; then + log "multiop failed $RC1" + else + kill -USR1 $pid + sleep $(( $TIMEOUT+1 )) + kill -0 $pid + [ $? -ne 0 ] && error "process isn't sleep" + start_ost || error "Unable to start OST1" + wait $pid || error "statfs failed" + fi + + umount_client $MOUNT || error "Unable to unmount client" + stop_ost || error "Unable to stop OST1" + stop_mds || error "Unable to stop MDS" +} +run_test 50e "normal statfs all servers down ==========================" + +test_50f() { + local RC1 + local pid + CONN_PROC="osc.$FSNAME-OST0001-osc-[M]*.ost_server_uuid" + + start_mds || error "Unable to start mds" + #first client should see only one ost + start_ost || error "Unable to start OST1" + wait_osc_import_state mds ost FULL + + start_ost2 || error "Unable to start OST2" + wait_osc_import_state mds ost2 FULL + + # Wait for client to detect down OST + stop_ost2 || error "Unable to stop OST2" + + wait_osc_import_state mds ost2 DISCONN + mount_client $MOUNT || error "Unable to mount client" + lctl set_param llite.$FSNAME-*.lazystatfs=0 + + multiop_bg_pause $MOUNT _f + RC1=$? + pid=$! + + if [ $RC1 -ne 0 ]; then + log "lazystatfs multiop failed $RC1" + else + kill -USR1 $pid + sleep $(( $TIMEOUT+1 )) + kill -0 $pid + [ $? -ne 0 ] && error "process isn't sleep" + start_ost2 || error "Unable to start OST2" + wait $pid || error "statfs failed" + stop_ost2 || error "Unable to stop OST2" + fi + + umount_client $MOUNT || error "Unable to unmount client" + stop_ost || error "Unable to stop OST1" + stop_mds || error "Unable to stop MDS" + writeconf +} +run_test 50f "normal statfs one server in down ==========================" + +test_50g() { + [ "$OSTCOUNT" -lt "2" ] && skip_env "$OSTCOUNT < 2, skipping" && return + setup + start_ost2 || error "Unable to start OST2" + + local PARAM="${FSNAME}-OST0001.osc.active" + + $LFS setstripe -c -1 $DIR/$tfile || error "Unable to lfs setstripe" + do_facet mgs $LCTL conf_param $PARAM=0 || error "Unable to deactivate OST" + + umount_client $MOUNT || error "Unable to unmount client" + mount_client $MOUNT || error "Unable to mount client" + # This df should not cause a panic + df -k $MOUNT + + do_facet mgs $LCTL conf_param $PARAM=1 || error "Unable to activate OST" + rm -f $DIR/$tfile + umount_client $MOUNT || error "Unable to unmount client" + stop_ost2 || error "Unable to stop OST2" + stop_ost || error "Unable to stop OST1" + stop_mds || error "Unable to stop MDS" + writeconf +} +run_test 50g "deactivated OST should not cause panic=====================" + +test_51() { + local LOCAL_TIMEOUT=20 + + reformat + setup_noconfig + check_mount || return 1 + + mkdir $MOUNT/d1 + $LFS setstripe -c -1 $MOUNT/d1 + #define OBD_FAIL_MDS_REINT_DELAY 0x142 + do_facet $SINGLEMDS "lctl set_param fail_loc=0x142" + touch $MOUNT/d1/f1 & + local pid=$! + sleep 2 + start_ost2 || return 2 + wait $pid + stop_ost2 || return 3 + cleanup +} +run_test 51 "Verify that mdt_reint handles RMF_MDT_MD correctly when an OST is added" + +copy_files_xattrs() +{ + local node=$1 + local dest=$2 + local xattrs=$3 + shift 3 + + do_node $node mkdir -p $dest + [ $? -eq 0 ] || { error "Unable to create directory"; return 1; } + + do_node $node 'tar cf - '$@' | tar xf - -C '$dest'; + [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1' + [ $? -eq 0 ] || { error "Unable to tar files"; return 2; } + + do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs + [ $? -eq 0 ] || { error "Unable to read xattrs"; return 3; } +} + +diff_files_xattrs() +{ + local node=$1 + local backup=$2 + local xattrs=$3 + shift 3 + + local backup2=${TMP}/backup2 + + do_node $node mkdir -p $backup2 + [ $? -eq 0 ] || { error "Unable to create directory"; return 1; } + + do_node $node 'tar cf - '$@' | tar xf - -C '$backup2'; + [ \"\${PIPESTATUS[*]}\" = \"0 0\" ] || exit 1' + [ $? -eq 0 ] || { error "Unable to tar files to diff"; return 2; } + + do_node $node "diff -rq $backup $backup2" + [ $? -eq 0 ] || { error "contents differ"; return 3; } + + local xattrs2=${TMP}/xattrs2 + do_node $node 'getfattr -d -m "[a-z]*\\." '$@' > '$xattrs2 + [ $? -eq 0 ] || { error "Unable to read xattrs to diff"; return 4; } + + do_node $node "diff $xattrs $xattrs2" + [ $? -eq 0 ] || { error "xattrs differ"; return 5; } + + do_node $node "rm -rf $backup2 $xattrs2" + [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 6; } +} + +test_52() { + start_mds + [ $? -eq 0 ] || { error "Unable to start MDS"; return 1; } + start_ost + [ $? -eq 0 ] || { error "Unable to start OST1"; return 2; } + mount_client $MOUNT + [ $? -eq 0 ] || { error "Unable to mount client"; return 3; } + + local nrfiles=8 + local ost1mnt=${MOUNT%/*}/ost1 + local ost1node=$(facet_active_host ost1) + local ost1tmp=$TMP/conf52 + + mkdir -p $DIR/$tdir + [ $? -eq 0 ] || { error "Unable to create tdir"; return 4; } + touch $TMP/modified_first + [ $? -eq 0 ] || { error "Unable to create temporary file"; return 5; } + do_node $ost1node "mkdir -p $ost1tmp && touch $ost1tmp/modified_first" + [ $? -eq 0 ] || { error "Unable to create temporary file"; return 6; } + sleep 1 + + $LFS setstripe $DIR/$tdir -c -1 -s 1M + [ $? -eq 0 ] || { error "lfs setstripe failed"; return 7; } + + for (( i=0; i < nrfiles; i++ )); do + multiop $DIR/$tdir/$tfile-$i Ow1048576w1048576w524288c + [ $? -eq 0 ] || { error "multiop failed"; return 8; } + echo -n . + done + echo + + # backup files + echo backup files to $TMP/files + local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first) + copy_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files + [ $? -eq 0 ] || { error "Unable to copy files"; return 9; } + + umount_client $MOUNT + [ $? -eq 0 ] || { error "Unable to umount client"; return 10; } + stop_ost + [ $? -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 + [ $? -eq 0 ] || { error "Unable to mount ost1 as ldiskfs"; return 12; } + + # backup objects + echo backup objects to $ost1tmp/objects + local objects=$(do_node $ost1node 'find '$ost1mnt'/O/0 -type f -size +0'\ + '-newer '$ost1tmp'/modified_first -regex ".*\/[0-9]+"') + copy_files_xattrs $ost1node $ost1tmp/objects $ost1tmp/object_xattrs $objects + [ $? -eq 0 ] || { error "Unable to copy objects"; return 13; } + + # move objects to lost+found + do_node $ost1node 'mv '$objects' '${ost1mnt}'/lost+found' + [ $? -eq 0 ] || { error "Unable to move objects"; return 14; } + + # recover objects + do_node $ost1node "ll_recover_lost_found_objs -d $ost1mnt/lost+found" + [ $? -eq 0 ] || { error "ll_recover_lost_found_objs failed"; return 15; } + + # compare restored objects against saved ones + 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" + [ $? -eq 0 ] || { error "Unable to umount ost1 as ldiskfs"; return 17; } + + start_ost + [ $? -eq 0 ] || { error "Unable to start ost1"; return 18; } + mount_client $MOUNT + [ $? -eq 0 ] || { error "Unable to mount client"; return 19; } + + # compare files + diff_files_xattrs `hostname` $TMP/files $TMP/file_xattrs $files + [ $? -eq 0 ] || { error "Unable to diff files"; return 20; } + + rm -rf $TMP/files $TMP/file_xattrs + [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 21; } + do_node $ost1node "rm -rf $ost1tmp" + [ $? -eq 0 ] || { error "Unable to delete temporary files"; return 22; } + cleanup +} +run_test 52 "check recovering objects from lost+found" + +# Checks threads_min/max/started for some service +# +# Arguments: service name (OST or MDT), facet (e.g., ost1, $SINGLEMDS), and a +# parameter pattern prefix like 'ost.*.ost'. +thread_sanity() { + local modname=$1 + local facet=$2 + local parampat=$3 + local opts=$4 + local tmin + local tmin2 + local tmax + local tmax2 + local tstarted + local paramp + local msg="Insane $modname thread counts" + shift 4 + + setup + check_mount || return 41 + + # We need to expand $parampat, but it may match multiple parameters, so + # we'll pick the first one + if ! paramp=$(do_facet $facet "lctl get_param -N ${parampat}.threads_min"|head -1); then + error "Couldn't expand ${parampat}.threads_min parameter name" + return 22 + fi + + # Remove the .threads_min part + paramp=${paramp%.threads_min} + + # Check for sanity in defaults + tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0) + tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0) + tstarted=$(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 $? + + # Check that we can lower min/max + do_facet $facet "lctl set_param ${paramp}.threads_min=$((tmin - 1))" + do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmax - 10))" + 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 -10)))' || return $? + + # Check that we can set min/max to the same value + do_facet $facet "lctl set_param ${paramp}.threads_min=$tmin" + do_facet $facet "lctl set_param ${paramp}.threads_max=$tmin" + tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0) + tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0) + lassert 26 "$msg" '(($tmin2 == $tmin && $tmax2 == $tmin))' || return $? + + # Check that we can't set max < min + do_facet $facet "lctl set_param ${paramp}.threads_max=$((tmin - 1))" + tmin2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min" || echo 0) + tmax2=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max" || echo 0) + lassert 27 "$msg" '(($tmin <= $tmax2))' || return $? + + # We need to ensure that we get the module options desired; to do this + # we set LOAD_MODULES_REMOTE=true and we call setmodopts below. + LOAD_MODULES_REMOTE=true + cleanup + local oldvalue + setmodopts -a $modname "$opts" oldvalue + + load_modules + setup + check_mount || return 41 + + # Restore previous setting of MODOPTS_* + setmodopts $modname "$oldvalue" + + # Check that $opts took + tmin=$(do_facet $facet "lctl get_param -n ${paramp}.threads_min") + tmax=$(do_facet $facet "lctl get_param -n ${paramp}.threads_max") + tstarted=$(do_facet $facet "lctl get_param -n ${paramp}.threads_started") + lassert 28 "$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' +} +run_test 53a "check OSS thread count params" + +test_53b() { + thread_sanity MDT $SINGLEMDS 'mdt.*.*.' 'mdt_num_threads=64' +} +run_test 53b "check MDT thread count params" + +if ! combined_mgs_mds ; then + stop mgs +fi + +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) + [ $? -eq 0 ] || error "llverdev failed!" + reformat_and_config +} +run_test 54a "llverdev" + +test_54b() { + setup + run_llverfs $MOUNT + [ $? -eq 0 ] || error "llverfs failed!" + cleanup +} +run_test 54b "llverfs" + +lov_objid_size() +{ + local max_ost_index=$1 + echo -n $(((max_ost_index + 1) * 8)) +} + +test_55() { + local mdsdev=$(mdsdevname 1) + local ostdev=$(ostdevname 1) + local saved_opts=$OST_MKFS_OPTS + + for i in 0 1023 2048 + do + OST_MKFS_OPTS="$saved_opts --index $i" + reformat + + setup_noconfig + stopall + + setup + 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 + error "lov_objid size has to be $(lov_objid_size $i), not $LOV_OBJID_SIZE" + else + echo ok, lov_objid size is correct: $LOV_OBJID_SIZE + fi + stopall + done + + OST_MKFS_OPTS=$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) + + start_mds + start_ost + start_ost2 || error "Unable to start second ost" + mount_client $MOUNT || error "Unable to mount client" + echo ok + $LFS osts + stopall + reformat +} +run_test 56 "check big indexes" cleanup_gss equals_msg `basename $0`: test complete