X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Ftests%2Fconf-sanity.sh;h=48252bf22613f9abb878fa142c13d23aa1032db6;hp=60f80d973109f746a4b29fd861ee56eecd806d59;hb=36e5b7203d5f80b5ac1341a05a35c19915c05bd6;hpb=87b0e9a2af342d661348fb4d4394b53dd0d94380;ds=sidebyside diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh old mode 100644 new mode 100755 index 60f80d9..48252bf --- a/lustre/tests/conf-sanity.sh +++ b/lustre/tests/conf-sanity.sh @@ -86,8 +86,9 @@ init_logging # require_dsh_mds || exit 0 require_dsh_ost || exit 0 -# -[ "$SLOW" = "no" ] && EXCEPT_SLOW="30a 31 45 69" + +# 8 22 (min)" +[ "$SLOW" = "no" ] && EXCEPT_SLOW="45 69" assert_DIR @@ -2024,7 +2025,8 @@ t32_test() { pushd $tmp/mnt/lustre fi $r cat $list_file | sort -k 6 >$tmp/list.orig - ls -Rni --time-style=+%s | sort -k 6 >$tmp/list || { + ls -Rni --time-style=+%s | sort -k 6 | + sed 's/\. / /' >$tmp/list || { error_noexit "ls" return 1 } @@ -2991,7 +2993,7 @@ test_43a() { [ $UID -ne 0 -o $RUNAS_ID -eq 0 ] && skip_env "run as root" ID1=${ID1:-501} - USER1=$(cat /etc/passwd | grep :$ID1:$ID1: | cut -d: -f1) + USER1=$(getent passwd | grep :$ID1:$ID1: | cut -d: -f1) [ -z "$USER1" ] && skip_env "missing user with uid=$ID1 gid=$ID1" && return @@ -3197,8 +3199,8 @@ test_45() { #17310 df -h $MOUNT & log "sleep 60 sec" sleep 60 - #define OBD_FAIL_PTLRPC_LONG_UNLINK 0x50f - do_facet client "$LCTL set_param fail_loc=0x50f" +#define OBD_FAIL_PTLRPC_LONG_REPL_UNLINK 0x50f + do_facet client "$LCTL set_param fail_loc=0x50f fail_val=0" log "sleep 10 sec" sleep 10 manual_umount_client --force || error "manual_umount_client failed" @@ -3810,9 +3812,9 @@ test_52() { echo # backup files - echo backup files to $TMP/files + echo backup files to $TMP/$tdir local files=$(find $DIR/$tdir -type f -newer $TMP/modified_first) - copy_files_xattrs $(hostname) $TMP/files $TMP/file_xattrs $files || + copy_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files || error "Unable to copy files" umount_client $MOUNT || error "Unable to umount client" @@ -3852,10 +3854,10 @@ test_52() { error "Some entry under /lost+found should be repaired" # compare files - diff_files_xattrs $(hostname) $TMP/files $TMP/file_xattrs $files || + diff_files_xattrs $(hostname) $TMP/$tdir $TMP/file_xattrs $files || error "Unable to diff files" - rm -rf $TMP/files $TMP/file_xattrs || + rm -rf $TMP/$tdir $TMP/file_xattrs || error "Unable to delete temporary files" do_node $ost1node "rm -rf $ost1tmp" || error "Unable to delete temporary files" @@ -4752,6 +4754,59 @@ test_70d() { } run_test 70d "stop MDT1, mkdir succeed, create remote dir fail" +test_70e() { + [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return + + [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.62) ] || + { skip "Need MDS version at least 2.7.62"; return 0; } + + cleanup || error "cleanup failed with $?" + + local mdsdev=$(mdsdevname 1) + local ostdev=$(ostdevname 1) + local mdsvdev=$(mdsvdevname 1) + local ostvdev=$(ostvdevname 1) + local opts_mds="$(mkfs_opts mds1 $mdsdev) --reformat $mdsdev $mdsvdev" + local opts_ost="$(mkfs_opts ost1 $ostdev) --reformat $ostdev $ostvdev" + + add mds1 $opts_mds || error "add mds1 failed" + start_mdt 1 || error "start mdt1 failed" + add ost1 $opts_ost || error "add ost1 failed" + start_ost || error "start ost failed" + mount_client $MOUNT > /dev/null || error "mount client $MOUNT failed" + + local soc=$(do_facet mds1 "$LCTL get_param -n \ + mdt.*MDT0000.sync_lock_cancel") + [ $soc == "never" ] || error "SoC enabled on single MDS" + + for i in $(seq 2 $MDSCOUNT); do + mdsdev=$(mdsdevname $i) + mdsvdev=$(mdsvdevname $i) + opts_mds="$(mkfs_opts mds$i $mdsdev) --reformat $mdsdev \ + $mdsvdev" + add mds$i $opts_mds || error "add mds$i failed" + start_mdt $i || error "start mdt$i fail" + done + + wait_dne_interconnect + + for i in $(seq $MDSCOUNT); do + soc=$(do_facet mds$i "$LCTL get_param -n \ + mdt.*MDT000$((i - 1)).sync_lock_cancel") + [ $soc == "blocking" ] || error "SoC not enabled on DNE" + done + + for i in $(seq 2 $MDSCOUNT); do + stop_mdt $i || error "stop mdt$i fail" + done + soc=$(do_facet mds1 "$LCTL get_param -n \ + mdt.*MDT0000.sync_lock_cancel") + [ $soc == "never" ] || error "SoC enabled on single MDS" + + cleanup || error "cleanup failed with $?" +} +run_test 70e "Sync-on-Cancel will be enabled by default on DNE" + test_71a() { [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return if combined_mgs_mds; then @@ -5062,6 +5117,28 @@ test_76b() { # LU-4783 } run_test 76b "verify params log setup correctly" +test_76c() { + [[ $(lustre_version_code mgs) -ge $(version_code 2.8.54) ]] || + { skip "Need MDS version at least 2.4.52" && return 0; } + setupall + local MASK_PARAM="mdd.*.changelog_mask" + echo "Change changelog_mask" + do_facet mgs $LCTL set_param -P $MASK_PARAM=-CLOSE || + error "Can't change changlog_mask" + wait_update $(facet_host mds) "$LCTL get_param -n $MASK_PARAM | + grep 'CLOSE'" "" + + echo "Check the value is stored after mds remount" + stop_mds || error "Failed to stop MDS" + start_mds || error "Failed to start MDS" + local CHANGELOG_MASK=$(do_facet mgs $LCTL get_param -n $MASK_PARAM) + echo $CHANGELOG_MASK | grep CLOSE > /dev/null && + error "changelog_mask is not changed" + + stopall +} +run_test 76c "verify changelog_mask is applied with set_param -P" + test_77() { # LU-3445 local server_version=$(lustre_version_code $SINGLEMDS) @@ -5109,8 +5186,12 @@ test_78() { skip "only applicable to ldiskfs-based MDTs and OSTs" && return # reformat the Lustre filesystem with a smaller size + local saved_MDSCOUNT=$MDSCOUNT local saved_MDSSIZE=$MDSSIZE + local saved_OSTCOUNT=$OSTCOUNT local saved_OSTSIZE=$OSTSIZE + MDSCOUNT=1 + OSTCOUNT=1 MDSSIZE=$((MDSSIZE - 20000)) OSTSIZE=$((OSTSIZE - 20000)) reformat || error "(1) reformat Lustre filesystem failed" @@ -5125,11 +5206,26 @@ test_78() { local i local file local num_files=100 + mkdir $MOUNT/$tdir || error "(3) mkdir $MOUNT/$tdir failed" + $LFS df; $LFS df -i for i in $(seq $num_files); do file=$MOUNT/$tdir/$tfile-$i - dd if=/dev/urandom of=$file count=1 bs=1M || + dd if=/dev/urandom of=$file count=1 bs=1M || { + $LCTL get_param osc.*.cur*grant* + $LFS df; $LFS df -i; + # stop creating files if there is no more space + if [ ! -e $file ]; then + num_files=$((i - 1)) + break + fi + + $LFS getstripe -v $file + local ost_idx=$(LFS getstripe -i $file) + do_facet ost$((ost_idx + 1)) \ + $LCTL get_param obdfilter.*.*grant* error "(4) create $file failed" + } done # unmount the Lustre filesystem @@ -5241,6 +5337,9 @@ test_78() { # unmount and reformat the Lustre filesystem cleanup || error "(12) cleanup Lustre filesystem failed" combined_mgs_mds || stop_mgs || error "(13) stop mgs failed" + + MDSCOUNT=$saved_MDSCOUNT + OSTCOUNT=$saved_OSTCOUNT reformat || error "(14) reformat Lustre filesystem failed" } run_test 78 "run resize2fs on MDT and OST filesystems" @@ -5428,8 +5527,9 @@ test_82a() { # LU-4665 local i local index local ost_indices + local LOV_V1_INSANE_STRIPE_COUNT=65532 for i in $(seq $OSTCOUNT); do - index=$((RANDOM * 2)) + index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT)) ost_indices+=" $index" done ost_indices=$(comma_list $ost_indices) @@ -5535,8 +5635,9 @@ test_82b() { # LU-4665 local i local index local ost_indices + local LOV_V1_INSANE_STRIPE_COUNT=65532 for i in $(seq $OSTCOUNT); do - index=$((RANDOM * 2)) + index=$(((RANDOM * 2) % LOV_V1_INSANE_STRIPE_COUNT)) ost_indices+=" $index" done ost_indices=$(comma_list $ost_indices) @@ -5664,32 +5765,6 @@ test_83() { run_test 83 "ENOSPACE on OST doesn't cause message VFS: \ Busy inodes after unmount ..." -recovery_time_min() { - local CONNECTION_SWITCH_MIN=5 - local CONNECTION_SWITCH_INC=5 - local CONNECTION_SWITCH_MAX - local RECONNECT_DELAY_MAX - local INITIAL_CONNECT_TIMEOUT - local max - local TO_20 - - #CONNECTION_SWITCH_MAX=min(50, max($CONNECTION_SWITCH_MIN,$TIMEOUT) - (($CONNECTION_SWITCH_MIN>$TIMEOUT)) && \ - max=$CONNECTION_SWITCH_MIN || max=$TIMEOUT - (($max<50)) && CONNECTION_SWITCH_MAX=$max || CONNECTION_SWITCH_MAX=50 - - #INITIAL_CONNECT_TIMEOUT = max(CONNECTION_SWITCH_MIN, \ - #obd_timeout/20) - TO_20=$(($TIMEOUT/20)) - (($CONNECTION_SWITCH_MIN>$TO_20)) && \ - INITIAL_CONNECT_TIMEOUT=$CONNECTION_SWITCH_MIN || \ - INITIAL_CONNECT_TIMEOUT=$TO_20 - - RECONNECT_DELAY_MAX=$(($CONNECTION_SWITCH_MAX+$CONNECTION_SWITCH_INC+ \ - $INITIAL_CONNECT_TIMEOUT)) - echo $((2*$RECONNECT_DELAY_MAX)) -} - test_84() { local facet=$SINGLEMDS local num=$(echo $facet | tr -d "mds") @@ -6263,6 +6338,140 @@ test_91() { } run_test 91 "evict-by-nid support" +generate_ldev_conf() { + # generate an ldev.conf file + local ldevconfpath=$1 + touch $ldevconfpath + printf "%s\t-\t%s-MGS0000\t%s\n" \ + $mgs_HOST \ + $FSNAME \ + $(mgsdevname) >> $ldevconfpath + + local mdsfo_host=$mdsfailover_HOST; + if [ -z "$mdsfo_host" ]; then + mdsfo_host="-" + fi + + for num in $(seq $MDSCOUNT); do + printf "%s\t%s\t%s-MDT%04d\t%s\n" \ + $mds_HOST \ + $mdsfo_host \ + $FSNAME \ + $num \ + $(mdsdevname $num) >> $ldevconfpath + done + + local ostfo_host=$ostfailover_HOST; + if [ -z "$ostfo_host" ]; then + ostfo_host="-" + fi + + for num in $(seq $OSTCOUNT); do + printf "%s\t%s\t%s-OST%04d\t%s\n" \ + $ost_HOST \ + $ostfo_host \ + $FSNAME \ + $num \ + $(ostdevname $num) >> $ldevconfpath + done +} + +generate_nids() { + # generate a nids file (mapping between hostname to nid) + # looks like we only have the MGS nid available to us + # so just echo that to a file + local nidspath=$1 + touch $nidspath + echo -e "${mgs_HOST}\t${MGSNID}" >> $nidspath +} + +test_92() { + local LDEVCONFPATH=$TMP/ldev.conf + local NIDSPATH=$TMP/nids + + echo "Host is $(hostname)" + + generate_ldev_conf $LDEVCONFPATH + generate_nids $NIDSPATH + + echo "----- ldev.conf -----" + cat $LDEVCONFPATH + echo "--- END ldev.conf ---" + + echo "----- /etc/nids -----" + cat $NIDSPATH + echo "--- END /etc/nids ---" + + # ldev can be in our build tree and if we aren't in a + # build tree, use 'which' to try and find it + local LDEV=$LUSTRE/scripts/ldev + [ ! -f "$LDEV" ] && local LDEV=$(which ldev 2> /dev/null) + + echo "ldev path is $LDEV" + + if [ ! -f "$LDEV" ]; then + rm $LDEVCONFPATH $NIDSPATH + error "failed to find ldev!" + fi + + # echo the mgs nid and compare it to environment variable MGSNID + # also, ldev.conf and nids is a server side thing, use the OSS + # hostname + local output + output=$(perl $LDEV -c $LDEVCONFPATH -H \ + $ost_HOST -n $NIDSPATH echo %m) + + echo "-- START OF LDEV OUTPUT --" + echo -e "$output" + echo "--- END OF LDEV OUTPUT ---" + + # ldev failed, error + if [ $? -ne 0 ]; then + rm $LDEVCONFPATH $NIDSPATH + error "ldev failed to execute!" + fi + + # need to process multiple lines because of combined MGS and MDS + echo -e $output | awk '{ print $2 }' | while read -r line ; do + if [ "$line" != "$MGSNID" ]; then + rm $LDEVCONFPATH $NIDSPATH + error "ldev failed mgs nid '$line', expected '$MGSNID'" + fi + done + + rm $LDEVCONFPATH $NIDSPATH +} +run_test 92 "ldev returns MGS NID correctly in command substitution" + +test_93() { + [ $MDSCOUNT -lt 3 ] && skip "needs >= 3 MDTs" && return + + reformat + #start mgs or mgs/mdt0 + if ! combined_mgs_mds ; then + start_mgs + start_mdt 1 + else + start_mdt 1 + fi + + start_ost || error "OST0 start fail" + + #define OBD_FAIL_MGS_WRITE_TARGET_DELAY 0x90e + do_facet mgs "$LCTL set_param fail_val = 10 fail_loc=0x8000090e" + for num in $(seq 2 $MDSCOUNT); do + start_mdt $num & + done + + mount_client $MOUNT || error "mount client fails" + wait_osc_import_state mds ost FULL + wait_osc_import_state client ost FULL + check_mount || error "check_mount failed" + + cleanup || error "cleanup failed with $?" +} +run_test 93 "register mulitple MDT at the same time" + if ! combined_mgs_mds ; then stop mgs fi