}
umount_client() {
- local MOUNTPATH=$1
- echo "umount lustre on ${MOUNTPATH}....."
- zconf_umount $(hostname) $MOUNTPATH || return 97
+ local mountpath=$1
+ shift
+ echo "umount lustre on $mountpath....."
+ zconf_umount $HOSTNAME $mountpath $@ || return 97
}
manual_umount_client(){
local rc
local FORCE=$1
echo "manual umount lustre on ${MOUNT}...."
- do_facet client "umount -d ${FORCE} $MOUNT"
+ do_facet client "umount ${FORCE} $MOUNT"
rc=$?
return $rc
}
setup
touch $DIR/$tfile || error "touch $DIR/$tfile failed"
stop_ost || error "Unable to stop OST1"
- cleanup
+ umount_client $MOUNT -f || error “unmount $MOUNT failed”
+ cleanup_nocli
eno=$?
# ok for ost to fail shutdown
if [ 202 -ne $eno ] && [ 0 -ne $eno ]; then
# 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 -f $MOUNT &
UMOUNT_PID=$!
sleep 6
echo "killing umount"
start_mds || error "MDS start failed"
stop_ost || error "Unable to stop OST1"
mount_client $MOUNT || error "mount_client $MOUNT failed"
- cleanup || error "cleanup_nocli failed with $?"
+ umount_client $MOUNT -f || error "umount_client $MOUNT failed"
+ cleanup_nocli || error "cleanup_nocli failed with $?"
grep " $MOUNT " /etc/mtab &&
error "$MOUNT entry in mtab after unmount"
pass
# check_mount will block trying to contact ost
mcreate $DIR/$tfile || error "mcreate $DIR/$tfile failed"
rm -f $DIR/$tfile || error "remove $DIR/$tfile failed"
- umount_client $MOUNT
+ umount_client $MOUNT -f
pass
echo "Client mount with a running ost"
test_30b() {
setup
+ local orignids=$($LCTL get_param -n \
+ osc.$FSNAME-OST0000-osc-[^M]*.import | grep failover_nids)
+
+ local orignidcount=$(echo "$orignids" | wc -w)
+
# 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/$NEWVAL@/")
+ local OSTNID=$(do_facet ost1 "$LCTL get_param nis" | tail -1 | \
+ awk '{print $1}')
+ local ORIGVAL=$(echo $OSTNID | egrep -oi "[0-9]*@")
+ local NEWVAL=$((($(echo $ORIGVAL | egrep -oi "[0-9]*") + 20) % 256))
+ local NEW=$(echo $OSTNID | sed "s/$ORIGVAL/$NEWVAL@/")
echo "Using fake nid $NEW"
- TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
+ local TEST="$LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
grep failover_nids | sed -n 's/.*\($NEW\).*/\1/p'"
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 |
+ local NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
grep failover_nids)
echo $NIDS
- # The NIDS value is the failover nid strings and "[" and "]". So
- # we need to subtract the space taken by the delimiters. This has
- # changed from earlier version of Lustre but this test is run only
- # locally so this change will not break interop. See LU-3386
- NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
- echo "should have 2 failover nids: $NIDCOUNT"
- [ $NIDCOUNT -eq 2 ] || error "Failover nid not added"
+ local NIDCOUNT=$(echo "$NIDS" | wc -w)
+ echo "should have $((orignidcount + 1)) entries \
+ in failover nids string, have $NIDCOUNT"
+ [ $NIDCOUNT -eq $((orignidcount + 1)) ] ||
+ error "Failover nid not added"
+
do_facet mgs "$LCTL conf_param -d $FSNAME-OST0000.failover.node" ||
error "conf_param delete failed"
umount_client $MOUNT
NIDS=$($LCTL get_param -n osc.$FSNAME-OST0000-osc-[^M]*.import |
grep failover_nids)
echo $NIDS
- NIDCOUNT=$(($(echo "$NIDS" | wc -w) - 3))
- echo "only 1 final nid should remain: $NIDCOUNT"
- [ $NIDCOUNT -eq 1 ] || error "Failover nids not removed"
+ NIDCOUNT=$(echo "$NIDS" | wc -w)
+ echo "only $orignidcount final entries should remain \
+ in failover nids string, have $NIDCOUNT"
+ [ $NIDCOUNT -eq $orignidcount ] || error "Failover nids not removed"
cleanup || error "cleanup failed with rc $?"
}
local src=/etc/rc.d
local tmp=$TMP/t32_image_create
- if [ $FSNAME != t32fs -o $MDSCOUNT -ne 1 -o \
- \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o $OSTCOUNT -ne 1 -o \
- -z "$OSTDEV1" ]; then
- error "Needs FSNAME=t32fs MDSCOUNT=1 MDSDEV1=<nonexistent_file>" \
- "(or MDSDEV, in the case of b1_8) OSTCOUNT=1" \
- "OSTDEV1=<nonexistent_file>"
+ if [ $FSNAME != t32fs -o \( -z "$MDSDEV" -a -z "$MDSDEV1" \) -o \
+ $OSTCOUNT -ne 1 -o -z "$OSTDEV1" ]; then
+ error "Needs FSNAME=t32fs MDSCOUNT=2 " \
+ "MDSDEV1=<nonexistent_file>" \
+ "MDSDEV2=<nonexistent_file>" \
+ "(or MDSDEV, in the case of b1_8)" \
+ "OSTCOUNT=1 OSTDEV1=<nonexistent_file>"
fi
mkdir $tmp || {
setupall
pushd /mnt/$FSNAME
ls -Rni --time-style=+%s >$tmp/img/list
- find . ! -name .lustre -type f -exec sha1sum {} \; |
+ find ! -name .lustre -type f -exec sha1sum {} \; |
sort -k 2 >$tmp/img/sha1sums
popd
$LCTL get_param -n version | head -n 1 |
{ if (NF == 1) { getline } else { num++ } ; print $num;} }' \
| tr -d "*" > $tmp/img/ispace
+ if [ $MDSCOUNT -ge 2 ]; then
+ $LFS mkdir -i 1 /mnt/$FSNAME/remote_dir
+ tar cf - -C $tmp/src . | tar xf - -C /mnt/$FSNAME/remote_dir
+ fi
+
stopall
pushd $tmp/src
uname -m >$tmp/img/arch
mv ${MDSDEV1:-$MDSDEV} $tmp/img
+ for num in $(seq 2 $MDSCOUNT); do
+ local devname=$(mdsdevname $num)
+ mv $devname $tmp/img
+ done
mv $OSTDEV1 $tmp/img
version=$(sed -e 's/\(^[0-9]\+\.[0-9]\+\)\(.*$\)/\1/' $tmp/img/commit |
umount $tmp/mnt/lustre || rc=$?
fi
if $shall_cleanup_mdt; then
- $r umount -d $tmp/mnt/mdt || rc=$?
+ $r $UMOUNT $tmp/mnt/mdt || rc=$?
fi
if $shall_cleanup_mdt1; then
- $r umount -d $tmp/mnt/mdt1 || rc=$?
+ $r $UMOUNT $tmp/mnt/mdt1 || rc=$?
fi
if $shall_cleanup_ost; then
- $r umount -d $tmp/mnt/ost || rc=$?
+ $r $UMOUNT $tmp/mnt/ost || rc=$?
fi
$r rm -rf $tmp
local list
local fstype=$(facet_fstype $SINGLEMDS)
local mdt_dev=$tmp/mdt
+ local mdt2_dev=$tmp/mdt2
local ost_dev=$tmp/ost
+ local stripe_index
+ local dir
trap 'trap - RETURN; t32_test_cleanup' RETURN
load_modules
mkdir -p $tmp/mnt/lustre || error "mkdir $tmp/mnt/lustre failed"
- $r mkdir -p $tmp/mnt/{mdt,ost}
+ $r mkdir -p $tmp/mnt/{mdt,mdt1,ost}
$r tar xjvf $tarball -S -C $tmp || {
error_noexit "Unpacking the disk image tarball"
return 1
error_noexit "tunefs.lustre before mounting the MDT"
return 1
}
+
+ if $r test -f $mdt2_dev; then
+ $r $TUNEFS --dryrun $mdt2_dev || {
+ $r losetup -a
+ error_noexit "tunefs.lustre before mounting the MDT"
+ return 1
+ }
+ fi
+
if [ "$writeconf" ]; then
mopts=writeconf
if [ $fstype == "ldiskfs" ]; then
error_noexit "Enable mdt quota feature"
return 1
}
+ if $r test -f $mdt2_dev; then
+ $r $TUNEFS --quota $mdt2_dev || {
+ $r losetup -a
+ error_noexit "Enable mdt quota feature"
+ return 1
+ }
+ fi
fi
else
if [ -n "$($LCTL list_nids | grep -v '\(tcp\|lo\)[[:digit:]]*$')" ]; then
$r $MOUNT_CMD -o $mopts $mdt_dev $tmp/mnt/mdt
$r $LCTL replace_nids $fsname-OST0000 $ostnid
$r $LCTL replace_nids $fsname-MDT0000 $nid
- $r umount -d $tmp/mnt/mdt
+ $r $UMOUNT $tmp/mnt/mdt
fi
mopts=exclude=$fsname-OST0000
}
shall_cleanup_mdt=true
- if [ "$dne_upgrade" != "no" ]; then
+ if $r test -f $mdt2_dev; then
+ mopts=mgsnode=$nid,$mopts
+ $r $MOUNT_CMD -o $mopts $mdt2_dev $tmp/mnt/mdt1 || {
+ $r losetup -a
+ error_noexit "Mounting the MDT"
+ return 1
+ }
+
+ echo "mount new MDT....$mdt2_dev"
+ $r $LCTL set_param -n mdt.${fsname}*.enable_remote_dir=1 ||
+ error_noexit "enable remote dir create failed"
+
+ shall_cleanup_mdt1=true
+ elif [ "$dne_upgrade" != "no" ]; then
local fs2mdsdev=$(mdsdevname 1_2)
local fs2mdsvdev=$(mdsvdevname 1_2)
}
echo "mount new MDT....$fs2mdsdev"
- $r mkdir -p $tmp/mnt/mdt1
$r $MOUNT_CMD -o $mopts $fs2mdsdev $tmp/mnt/mdt1 || {
error_noexit "mount mdt1 failed"
return 1
error_noexit "Setting MDT1 \"lov.stripesize\""
return 1
}
-
fi
if [ "$writeconf" ]; then
error_noexit "Mounting the client"
return 1
}
+
shall_cleanup_lustre=true
- $LCTL set_param debug="$PTLDEBUG"
+ $r $LCTL set_param debug="$PTLDEBUG"
t32_verify_quota $node $fsname $tmp/mnt/lustre || {
error_noexit "verify quota failed"
}
if [ "$dne_upgrade" != "no" ]; then
- $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/remote_dir || {
+ $LFS mkdir -i 1 -c2 $tmp/mnt/lustre/striped_dir || {
error_noexit "set remote dir failed"
return 1
}
- $LFS setdirstripe -D -c2 $tmp/mnt/lustre/remote_dir
+ $LFS setdirstripe -D -c2 $tmp/mnt/lustre/striped_dir
pushd $tmp/mnt/lustre
- tar -cf - . --exclude=./remote_dir |
- tar -xvf - -C remote_dir 1>/dev/null || {
+ tar -cf - . --exclude=./striped_dir \
+ --exclude=./remote_dir |
+ tar -xvf - -C striped_dir 1>/dev/null || {
error_noexit "cp to remote dir failed"
return 1
}
popd
fi
+ # If it is upgrade from DNE (2.5), then rename the remote dir,
+ # which is created in 2.5 to striped dir.
+ if $r test -f $mdt2_dev; then
+ stripe_index=$(LFS getdirstripe -i \
+ $tmp/mnt/lustre/remote_dir)
+ [ $stripe_index -eq 1 ] || {
+ error_noexit "get index $striped_index failed"
+ return 1
+ }
+ mv $tmp/mnt/lustre/remote_dir \
+ $tmp/mnt/lustre/striped_dir/ || {
+ error_noexit "mv failed"
+ return 1
+ }
+ fi
+
dd if=/dev/zero of=$tmp/mnt/lustre/tmp_file bs=10k count=10 || {
error_noexit "dd failed"
return 1
# is identical
$r cat $tmp/sha1sums | sort -k 2 >$tmp/sha1sums.orig
if [ "$dne_upgrade" != "no" ]; then
- pushd $tmp/mnt/lustre/remote_dir
+ pushd $tmp/mnt/lustre/striped_dir
else
pushd $tmp/mnt/lustre
fi
- find ! -name .lustre -type f -exec sha1sum {} \; |
+ find ! -path "*remote_dir*" ! -name .lustre -type f \
+ -exec sha1sum {} \; |
sort -k 2 >$tmp/sha1sums || {
+ popd
error_noexit "sha1sum"
return 1
}
error_noexit "sha1sum verification failed"
return 1
fi
+
+ # if upgrade from DNE (2.5), then check remote directory
+ if $r test -f $mdt2_dev; then
+ pushd $tmp/mnt/lustre/striped_dir/remote_dir
+ find ! -name .lustre -type f \
+ -exec sha1sum {} \; |
+ sort -k 2 >$tmp/sha1sums || {
+ popd
+ error_noexit "sha1sum"
+ return 1
+ }
+ popd
+ if ! diff -ub $tmp/sha1sums.orig \
+ $tmp/sha1sums; then
+ error_noexit "sha1sum dne failed"
+ return 1
+ fi
+ fi
else
echo "sha1sum verification skipped"
fi
if [ "$dne_upgrade" != "no" ]; then
- rm -rf $tmp/mnt/lustre/remote_dir || {
+ rm -rf $tmp/mnt/lustre/striped_dir || {
error_noexit "remove remote dir failed"
return 1
}
echo "list verification skipped"
fi
+ # migrate files/dirs to remote MDT, then move them back
+ if [ $(lustre_version_code mds1) -ge $(version_code 2.7.50) -a \
+ $dne_upgrade != "no" ]; then
+ $r $LCTL set_param -n \
+ mdt.${fsname}*.enable_remote_dir=1 2>/dev/null
+
+ echo "test migration"
+ pushd $tmp/mnt/lustre
+ for dir in $(find ! -name .lustre ! -name . -type d); do
+ mdt_index=$($LFS getdirstripe -i $dir)
+ stripe_cnt=$($LFS getdirstripe -c $dir)
+ if [ $mdt_index = 0 -a $stripe_cnt -le 1 ]; then
+ $LFS mv -M 1 $dir || {
+ popd
+ error_noexit "migrate MDT1 failed"
+ return 1
+ }
+ fi
+ done
+
+ for dir in $(find ! -name . ! -name .lustre -type d); do
+ mdt_index=$($LFS getdirstripe -i $dir)
+ stripe_cnt=$($LFS getdirstripe -c $dir)
+ if [ $mdt_index = 1 -a $stripe_cnt -le 1 ]; then
+ $LFS mv -M 0 $dir || {
+ popd
+ error_noexit "migrate MDT0 failed"
+ return 1
+ }
+ fi
+ done
+ popd
+ fi
+
#
# When adding new data verification tests, please check for
# the presence of the required reference files first, like
shall_cleanup_lustre=false
else
if [ "$dne_upgrade" != "no" ]; then
- $r umount -d $tmp/mnt/mdt1 || {
+ $r $UMOUNT $tmp/mnt/mdt1 || {
error_noexit "Unmounting the MDT2"
return 1
}
shall_cleanup_mdt1=false
fi
- $r umount -d $tmp/mnt/mdt || {
+ $r $UMOUNT $tmp/mnt/mdt || {
error_noexit "Unmounting the MDT"
return 1
}
shall_cleanup_mdt=false
- $r umount -d $tmp/mnt/ost || {
+ $r $UMOUNT $tmp/mnt/ost || {
error_noexit "Unmounting the OST"
return 1
}
# Do not support 1_8 and 2_1 direct upgrade to DNE2 anymore */
echo $tarball | grep "1_8" && continue
echo $tarball | grep "2_1" && continue
+ load_modules
dne_upgrade=yes t32_test $tarball writeconf || rc=$?
done
return $rc
cp /etc/hosts $MOUNT2/ || error "copy /etc/hosts $MOUNT2/ failed"
$GETSTRIPE $MOUNT2/hosts || error "$GETSTRIPE $MOUNT2/hosts failed"
- umount -d $MOUNT2
+ umount $MOUNT2
stop fs2ost -f
stop fs2mds -f
cleanup_nocli || error "cleanup_nocli failed with $?"
rc=3
fi
- umount -d $MOUNT2
+ $UMOUNT $MOUNT2
stop fs3ost -f || error "unable to stop OST3"
stop fs2ost -f || error "unable to stop OST2"
stop fs2mds -f || error "unable to stop second MDS"
echo mount_op=$mount_op
- do_facet $SINGLEMDS "umount -d $mntpt && rm -f $mdsdev_sym"
+ do_facet $SINGLEMDS "$UMOUNT $mntpt && rm -f $mdsdev_sym"
if $(echo $mount_op | grep -q "unable to set tunable"); then
error "set tunables failed for symlink device"
# root should be able to access root file after that
#
local NIDLIST=$($LCTL list_nids all | tr '\n' ' ')
- NIDLIST="2@elan $NIDLIST 192.168.0.[2,10]@tcp"
+ NIDLIST="2@gni $NIDLIST 192.168.0.[2,10]@tcp"
NIDLIST=$(echo $NIDLIST | tr -s ' ' ' ')
set_conf_param_and_check mds \
"$LCTL get_param -n mdt.$FSNAME-MDT0000.nosquash_nids" \
stop_ost2 || error "Unable to stop OST2"
fi
- umount_client $MOUNT || error "Unable to unmount client"
+ umount_client $MOUNT -f || error "Unable to unmount client"
stop_ost || error "Unable to stop OST1"
stop_mds || error "Unable to stop MDS"
#writeconf to remove all ost2 traces for subsequent tests
}
run_test 50h "LU-642: activate deactivated OST"
+test_50i() {
+ # prepare MDT/OST, make OSC inactive for OST1
+ [ "$MDSCOUNT" -lt "2" ] && skip_env "$MDSCOUNT < 2, skipping" && return
+
+ [ $(facet_fstype ost1) == zfs ] && import_zpool ost1
+ load_modules
+ do_facet mds2 "$TUNEFS --param mdc.active=0 $(mdsdevname 2)" ||
+ error "tunefs MDT2 failed"
+ start_mds || error "Unable to start MDT"
+ start_ost || error "Unable to start OST1"
+ start_ost2 || error "Unable to start OST2"
+ mount_client $MOUNT || error "client start failed"
+
+ mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
+
+ $LCTL conf_param ${FSNAME}-MDT0000.mdc.active=0 &&
+ error "deactive MDC0 succeeds"
+ # activate MDC for MDT2
+ local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
+ set_conf_param_and_check client \
+ "$TEST" "${FSNAME}-MDT0001.mdc.active" 1 ||
+ error "Unable to activate MDT2"
+
+ wait_clients_import_state ${CLIENTS:-$HOSTNAME} mds2 FULL
+ if [ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.60) ]
+ then
+ wait_dne_interconnect
+ fi
+ $LFS mkdir -i1 $DIR/$tdir/2 || error "mkdir $DIR/$tdir/2 failed"
+ # create some file
+ createmany -o $DIR/$tdir/2/$tfile-%d 1 || error "create files failed"
+
+ rm -rf $DIR/$tdir/2 || error "unlink dir failed"
+
+ # deactivate MDC for MDT2
+ local TEST="$LCTL get_param -n mdc.${FSNAME}-MDT0001-mdc-[!M]*.active"
+ set_conf_param_and_check client \
+ "$TEST" "${FSNAME}-MDT0001.mdc.active" 0 ||
+ error "Unable to deactivate MDT2"
+
+ $LFS mkdir -i1 $DIR/$tdir/2 &&
+ error "mkdir $DIR/$tdir/2 succeeds after deactive MDT"
+
+ # cleanup
+ umount_client $MOUNT || error "Unable to umount client"
+ stop_mds
+ stop_ost
+ stop_ost 2
+}
+run_test 50i "activate deactivated MDT"
+
test_51() {
local LOCAL_TIMEOUT=20
start_ost2 || error "Unable to start OST1"
wait $pid
stop_ost2 || error "Unable to stop OST1"
- cleanup || error "cleanup failed with $?"
+ umount_client $MOUNT -f || error “unmount $MOUNT failed”
+ cleanup_nocli || error “stop server failed”
#writeconf to remove all ost2 traces for subsequent tests
writeconf_or_reformat
}
MDSJOURNALSIZE=16
for num in $(seq 1 $MDSCOUNT); do
- add mds${num} $(mkfs_opts mds${num} $(mdsdevname $num)) \
- --reformat $(mdsdevname $num) $(mdsvdevname $num)
+ reformat_mdt $num
done
add ost1 $(mkfs_opts ost1 $(ostdevname 1)) --index=10000 --reformat \
$(ostdevname 1) $(ostvdevname 1)
local large_value="$(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 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"
+ 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"
+ 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"
+ 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"
+ 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"
+ 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 "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"
+ log "remove large xattr $name from $file"
+ setfattr -x $name $file || error "removing $name from $file failed"
- rm -f $file
- stopall
+ if $lxattr; then
+ stopall || error "stopping for e2fsck run"
+ for num in $(seq $MDSCOUNT); do
+ run_e2fsck $(facet_active_host mds$num) \
+ $(mdsdevname $num) "-y" ||
+ error "e2fsck MDT$num failed"
+ done
+ setup_noconfig || error "remounting the filesystem failed"
+ fi
+
+ # need to delete this file to avoid problems in other tests
+ rm -f $file
+ stopall || error "stopping systems to turn off large_xattr"
if $lxattr; then
for num in $(seq $MDSCOUNT); do
do_facet mds${num} $TUNE2FS -O ^large_xattr \
stop_ost2 || error "Unable to stop second ost"
echo "$LFS df"
$LFS df --lazy || error "lfs df failed"
- cleanup || error "cleanup failed with $?"
+ umount_client $MOUNT -f || error “unmount $MOUNT failed”
+ cleanup_nocli || error "cleanup_nocli failed with $?"
#writeconf to remove all ost2 traces for subsequent tests
writeconf_or_reformat
}
do_facet $SINGLEMDS \
"mount -t $(facet_fstype $SINGLEMDS) $opts $devname $brpt"
do_facet $SINGLEMDS "rm -f ${brpt}/last_rcvd"
- do_facet $SINGLEMDS "umount -d $brpt"
+ do_facet $SINGLEMDS "$UMOUNT $brpt"
# restart MDS, the "last_rcvd" file should be recreated.
start_mds || error "fail to restart the MDS"
}
run_test 68 "be able to reserve specific sequences in FLDB"
+# Test 69: is about the total number of objects ever created on an OST.
+# so that when it is reformatted the normal MDS->OST orphan recovery won't
+# just "precreate" the missing objects. In the past it might try to recreate
+# millions of objects after an OST was reformatted
test_69() {
local server_version=$(lustre_version_code $SINGLEMDS)
skip "Need MDS version at least 2.5.0" && return
setup
+ mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
# use OST0000 since it probably has the most creations
local OSTNAME=$(ostname_from_index 0)
local mdtosc_proc1=$(get_mdtosc_proc_path mds1 $OSTNAME)
+ local last_id=$(do_facet mds1 $LCTL get_param -n \
+ osc.$mdtosc_proc1.prealloc_last_id)
- # Want to have OST LAST_ID over 1.5 * OST_MAX_PRECREATE to
- # verify that the LAST_ID recovery is working properly. If
+ # Want to have OST LAST_ID over 5 * OST_MAX_PRECREATE to
+ # verify that the LAST_ID recovery is working properly. If
# not, then the OST will refuse to allow the MDS connect
# because the LAST_ID value is too different from the MDS
#define OST_MAX_PRECREATE=20000
- local num_create=$((20000 * 3))
+ local ost_max_pre=20000
+ local num_create=$(( ost_max_pre * 5 + 1 - last_id))
+
+ # If the LAST_ID is already over 5 * OST_MAX_PRECREATE, we don't
+ # need to create any files. So, skip this section.
+ if [ $num_create -gt 0 ]; then
+ # Check the number of inodes available on OST0
+ local files=0
+ local ifree=$($LFS df -i $MOUNT | awk '/OST0000/ { print $4 }')
+ log "On OST0, $ifree inodes available. Want $num_create."
+
+ $SETSTRIPE -i 0 $DIR/$tdir ||
+ error "$SETSTRIPE -i 0 $DIR/$tdir failed"
+ if [ $ifree -lt 10000 ]; then
+ files=$(( ifree - 50 ))
+ else
+ files=10000
+ fi
+
+ local j=$((num_create / files + 1))
+ for i in $(seq 1 $j); do
+ createmany -o $DIR/$tdir/$tfile-$i- $files ||
+ error "createmany fail create $files files: $?"
+ unlinkmany $DIR/$tdir/$tfile-$i- $files ||
+ error "unlinkmany failed unlink $files files"
+ done
+ fi
- mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
- $SETSTRIPE -i 0 $DIR/$tdir || error "$SETSTRIPE -i 0 $DIR/$tdir failed"
- createmany -o $DIR/$tdir/$tfile- $num_create ||
- error "createmany: failed to create $num_create files: $?"
# delete all of the files with objects on OST0 so the
# filesystem is not inconsistent later on
- $LFS find $MOUNT --ost 0 | xargs rm
+ $LFS find $MOUNT --ost 0 -print0 | xargs -0 rm
umount_client $MOUNT || error "umount client failed"
stop_ost || error "OST0 stop failure"
local idx=$($GETSTRIPE -i $DIR/$tdir/$tfile-last)
[ $idx -ne 0 ] && error "$DIR/$tdir/$tfile-last on $idx not 0" || true
+ local iused=$($LFS df -i $MOUNT | awk '/OST0000/ { print $3 }')
+ log "On OST0, $iused used inodes"
+ [ $iused -ge $((ost_max_pre/2 + 1000)) ] &&
+ error "OST replacement created too many inodes; $iused"
cleanup || error "cleanup failed with $?"
}
run_test 69 "replace an OST with the same index"
local time_min=$(recovery_time_min)
local recovery_duration
local completed_clients
+ local correct_clients
local wrap_up=5
+ load_modules
echo "start mds service on $(facet_active_host $facet)"
- start $facet ${dev} $MDS_MOUNT_OPTS \
- "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
+ start_mds \
+ "-o recovery_time_hard=$time_min,recovery_time_soft=$time_min" $@ ||
error "start MDS failed"
start_ost
#define OBD_FAIL_TGT_REPLAY_DELAY 0x709 | FAIL_SKIP
do_facet $SINGLEMDS "lctl set_param fail_loc=0x20000709 fail_val=5"
- facet_failover $SINGLEMDS || error "failover: $?"
+ facet_failover --fsck $SINGLEMDS || error "failover: $?"
client_up
echo "recovery status"
completed_clients=$(do_facet $SINGLEMDS \
"$LCTL get_param -n mdt.$FSNAME-MDT0000.recovery_status" |
awk '/completed_clients/ { print $2 }')
- [ "$completed_clients" = "1/2" ] ||
- error "completed_clients != 1/2: $completed_clients"
+
+ correct_clients="$MDSCOUNT/$((MDSCOUNT+1))"
+ [ "$completed_clients" = "${correct_clients}" ] ||
+ error "$completed_clients != $correct_clients"
do_facet $SINGLEMDS "lctl set_param fail_loc=0"
umount_client $MOUNT1
}
run_test 87 "check if MDT inode can hold EAs with N stripes properly"
+# $1 test directory
+# $2 (optional) value of max_mod_rpcs_in_flight to set
+check_max_mod_rpcs_in_flight() {
+ local dir="$1"
+ local mmr="$2"
+ local idx
+ local facet
+ local tmp
+ local i
+
+ idx=$(printf "%04x" $($LFS getdirstripe -i $dir))
+ facet="mds$((0x$idx + 1))"
+
+ if [ -z "$mmr" ]; then
+ # get value of max_mod_rcps_in_flight
+ mmr=$($LCTL get_param -n \
+ mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight) ||
+ error "Unable to get max_mod_rpcs_in_flight"
+ echo "max_mod_rcps_in_flight is $mmr"
+ else
+ # set value of max_mod_rpcs_in_flight
+ $LCTL set_param \
+ mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight=$mmr ||
+ error "Unable to set max_mod_rpcs_in_flight to $mmr"
+ echo "max_mod_rpcs_in_flight set to $mmr"
+ fi
+
+ # create mmr+1 files
+ echo "creating $((mmr + 1)) files ..."
+ umask 0022
+ for i in $(seq $((mmr + 1))); do
+ touch $dir/file-$i
+ done
+
+ ### part 1 ###
+
+ # consumes mmr-1 modify RPC slots
+ #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
+ # drop requests on MDT so that RPC slots are consumed
+ # during all the request resend interval
+ do_facet $facet "$LCTL set_param fail_loc=0x159"
+ echo "launch $((mmr - 1)) chmod in parallel ..."
+ for i in $(seq $((mmr - 1))); do
+ chmod 0600 $dir/file-$i &
+ done
+ sleep 1
+
+ # send one additional modify RPC
+ do_facet $facet "$LCTL set_param fail_loc=0"
+ echo "launch 1 additional chmod in parallel ..."
+ chmod 0600 $dir/file-$mmr &
+ sleep 1
+
+ # check this additional modify RPC get a modify RPC slot
+ # and succeed its operation
+ checkstat -vp 0600 $dir/file-$mmr ||
+ error "Unable to send $mmr modify RPCs in parallel"
+ wait
+
+ ### part 2 ###
+
+ # consumes mmr modify RPC slots
+ #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
+ # drop requests on MDT so that RPC slots are consumed
+ # during all the request resend interval
+ do_facet $facet "$LCTL set_param fail_loc=0x159"
+ echo "launch $mmr chmod in parallel ..."
+ for i in $(seq $mmr); do
+ chmod 0666 $dir/file-$i &
+ done
+ sleep 1
+
+ # send one additional modify RPC
+ do_facet $facet "$LCTL set_param fail_loc=0"
+ echo "launch 1 additional chmod in parallel ..."
+ chmod 0666 $dir/file-$((mmr + 1)) &
+ sleep 1
+
+ # check this additional modify RPC blocked getting a modify RPC slot
+ checkstat -vp 0644 $dir/file-$((mmr + 1)) ||
+ error "Unexpectedly send $(($mmr + 1)) modify RPCs in parallel"
+ wait
+}
+
+test_90a() {
+ reformat
+ if ! combined_mgs_mds ; then
+ start_mgs
+ fi
+ setup
+
+ [[ $($LCTL get_param mdc.*.import |
+ grep "connect_flags:.*multi_mod_rpc") ]] ||
+ { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
+
+ # check default value
+ $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
+ check_max_mod_rpcs_in_flight $DIR/$tdir
+
+ cleanup
+}
+run_test 90a "check max_mod_rpcs_in_flight is enforced"
+
+test_90b() {
+ local idx
+ local facet
+ local tmp
+ local mmrpc
+
+ setup
+
+ [[ $($LCTL get_param mdc.*.import |
+ grep "connect_flags:.*multi_mod_rpc") ]] ||
+ { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
+
+ ### test 1.
+ # update max_mod_rpcs_in_flight
+ $LFS mkdir -c1 $DIR/${tdir}1 || error "mkdir $DIR/${tdir}1 failed"
+ check_max_mod_rpcs_in_flight $DIR/${tdir}1 1
+
+ ### test 2.
+ # check client is able to send multiple modify RPCs in paralell
+ tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
+ grep -c "multi_mod_rpcs")
+ if [ "$tmp" -ne $MDSCOUNT ]; then
+ echo "Client not able to send multiple modify RPCs in parallel"
+ cleanup
+ return
+ fi
+
+ # update max_mod_rpcs_in_flight
+ $LFS mkdir -c1 $DIR/${tdir}2 || error "mkdir $DIR/${tdir}2 failed"
+ check_max_mod_rpcs_in_flight $DIR/${tdir}2 5
+
+ ### test 3.
+ $LFS mkdir -c1 $DIR/${tdir}3 || error "mkdir $DIR/${tdir}3 failed"
+ idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/${tdir}3))
+ facet="mds$((0x$idx + 1))"
+
+ # save MDT max_mod_rpcs_per_client
+ mmrpc=$(do_facet $facet \
+ cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
+
+ # update max_mod_rpcs_in_flight
+ umount_client $MOUNT
+ do_facet $facet \
+ "echo 16 > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
+ mount_client $MOUNT
+ $LCTL set_param mdc.$FSNAME-MDT$idx-mdc-*.max_rpcs_in_flight=17
+ check_max_mod_rpcs_in_flight $DIR/${tdir}3 16
+
+ # restore MDT max_mod_rpcs_per_client initial value
+ do_facet $facet \
+ "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
+
+ rm -rf $DIR/${tdir}?
+ cleanup
+}
+run_test 90b "check max_mod_rpcs_in_flight is enforced after update"
+
+test_90c() {
+ local tmp
+ local mrif
+ local mmrpc
+
+ setup
+
+ [[ $($LCTL get_param mdc.*.import |
+ grep "connect_flags:.*multi_mod_rpc") ]] ||
+ { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
+
+ # check client is able to send multiple modify RPCs in paralell
+ tmp=$($LCTL get_param -n mdc.$FSNAME-MDT*-mdc-*.import |
+ grep -c "multi_mod_rpcs")
+ if [ "$tmp" -ne $MDSCOUNT ]; then
+ skip "Client not able to send multiple modify RPCs in parallel"
+ cleanup
+ return
+ fi
+
+ # get max_rpcs_in_flight value
+ mrif=$($LCTL get_param -n mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight)
+ echo "max_rpcs_in_flight is $mrif"
+
+ # get MDT max_mod_rpcs_per_client
+ mmrpc=$(do_facet mds1 \
+ cat /sys/module/mdt/parameters/max_mod_rpcs_per_client)
+ echo "max_mod_rpcs_per_client is $mmrpc"
+
+ # testcase 1
+ # attempt to set max_mod_rpcs_in_flight to max_rpcs_in_flight value
+ # prerequisite: set max_mod_rpcs_per_client to max_rpcs_in_flight value
+ umount_client $MOUNT
+ do_facet mds1 \
+ "echo $mrif > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
+ mount_client $MOUNT
+
+ $LCTL set_param \
+ mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$mrif &&
+ error "set max_mod_rpcs_in_flight to $mrif should fail"
+
+ umount_client $MOUNT
+ do_facet mds1 \
+ "echo $mmrpc > /sys/module/mdt/parameters/max_mod_rpcs_per_client"
+ mount_client $MOUNT
+
+ # testcase 2
+ # attempt to set max_mod_rpcs_in_flight to max_mod_rpcs_per_client+1
+ # prerequisite: set max_rpcs_in_flight to max_mod_rpcs_per_client+2
+ $LCTL set_param \
+ mdc.$FSNAME-MDT0000-mdc-*.max_rpcs_in_flight=$((mmrpc + 2))
+
+ $LCTL set_param \
+ mdc.$FSNAME-MDT0000-mdc-*.max_mod_rpcs_in_flight=$((mmrpc + 1)) &&
+ error "set max_mod_rpcs_in_flight to $((mmrpc + 1)) should fail"
+
+ cleanup
+}
+run_test 90c "check max_mod_rpcs_in_flight update limits"
+
+test_90d() {
+ local idx
+ local facet
+ local mmr
+ local i
+ local pid
+
+ setup
+
+ [[ $($LCTL get_param mdc.*.import |
+ grep "connect_flags:.*multi_mod_rpc") ]] ||
+ { skip "Need MDC with 'multi_mod_rpcs' feature"; return 0; }
+
+ $LFS mkdir -c1 $DIR/$tdir || error "mkdir $DIR/$tdir failed"
+ idx=$(printf "%04x" $($LFS getdirstripe -i $DIR/$tdir))
+ facet="mds$((0x$idx + 1))"
+
+ # check client version supports multislots
+ tmp=$($LCTL get_param -N \
+ mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
+ if [ -z "$tmp" ]; then
+ skip "Client does not support multiple modify RPCs in flight"
+ cleanup
+ return
+ fi
+
+ # get current value of max_mod_rcps_in_flight
+ mmr=$($LCTL get_param -n \
+ mdc.$FSNAME-MDT$idx-mdc-*.max_mod_rpcs_in_flight)
+ echo "max_mod_rcps_in_flight is $mmr"
+
+ # create mmr files
+ echo "creating $mmr files ..."
+ umask 0022
+ for i in $(seq $mmr); do
+ touch $DIR/$tdir/file-$i
+ done
+
+ # prepare for close RPC
+ multiop_bg_pause $DIR/$tdir/file-close O_c
+ pid=$!
+
+ # consumes mmr modify RPC slots
+ #define OBD_FAIL_MDS_REINT_MULTI_NET 0x159
+ # drop requests on MDT so that RPC slots are consumed
+ # during all the request resend interval
+ do_facet $facet "$LCTL set_param fail_loc=0x159"
+ echo "launch $mmr chmod in parallel ..."
+ for i in $(seq $mmr); do
+ chmod 0600 $DIR/$tdir/file-$i &
+ done
+
+ # send one additional close RPC
+ do_facet $facet "$LCTL set_param fail_loc=0"
+ echo "launch 1 additional close in parallel ..."
+ kill -USR1 $pid
+ cancel_lru_locks mdc
+ sleep 1
+
+ # check this additional close RPC get a modify RPC slot
+ # and multiop process completed
+ [ -d /proc/$pid ] &&
+ error "Unable to send the additional close RPC in parallel"
+ wait
+ rm -rf $DIR/$tdir
+ cleanup
+}
+run_test 90d "check one close RPC is allowed above max_mod_rpcs_in_flight"
+
+check_uuid_on_ost() {
+ local nid=$1
+ do_facet ost1 "$LCTL get_param obdfilter.${FSNAME}*.exports.'$nid'.uuid"
+}
+
+check_uuid_on_mdt() {
+ local nid=$1
+ do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
+}
+
+test_91() {
+ local uuid
+ local nid
+ local found
+
+ load_modules
+
+ [[ $(lustre_version_code ost1) -ge $(version_code 2.7.63) ]] ||
+ { skip "Need OST version at least 2.7.63" && return 0; }
+ [[ $(lustre_version_code $SINGLEMDS) -ge $(version_code 2.7.63) ]] ||
+ { skip "Need MDT version at least 2.7.63" && return 0; }
+
+ start_mds || error "MDS start failed"
+ start_ost || error "unable to start OST"
+ mount_client $MOUNT || error "client start failed"
+ check_mount || error "check_mount failed"
+
+ if remote_mds; then
+ nid=$($LCTL list_nids | head -1 | sed "s/\./\\\./g")
+ else
+ nid="0@lo"
+ fi
+ uuid=$(get_client_uuid $MOUNT)
+
+ echo "list nids on mdt:"
+ do_facet $SINGLEMDS "$LCTL list_param mdt.${FSNAME}*.exports.*"
+ echo "uuid from $nid:"
+ do_facet $SINGLEMDS "$LCTL get_param mdt.${FSNAME}*.exports.'$nid'.uuid"
+
+ found=$(check_uuid_on_mdt $nid | grep $uuid)
+ [ -z "$found" ] && error "can't find $uuid $nid on MDT"
+ found=$(check_uuid_on_ost $nid | grep $uuid)
+ [ -z "$found" ] && error "can't find $uuid $nid on OST"
+
+ # umount the client so it won't reconnect
+ manual_umount_client --force || error "failed to umount $?"
+ # shouldn't disappear on MDS after forced umount
+ found=$(check_uuid_on_mdt $nid | grep $uuid)
+ [ -z "$found" ] && error "can't find $uuid $nid"
+
+ echo "evict $nid"
+ do_facet $SINGLEMDS \
+ "$LCTL set_param -n mdt.${mds1_svc}.evict_client nid:$nid"
+
+ found=$(check_uuid_on_mdt $nid | grep $uuid)
+ [ -n "$found" ] && error "found $uuid $nid on MDT"
+ found=$(check_uuid_on_ost $nid | grep $uuid)
+ [ -n "$found" ] && error "found $uuid $nid on OST"
+
+ # check it didn't reconnect (being umounted)
+ sleep $((TIMEOUT+1))
+ found=$(check_uuid_on_mdt $nid | grep $uuid)
+ [ -n "$found" ] && error "found $uuid $nid on MDT"
+ found=$(check_uuid_on_ost $nid | grep $uuid)
+ [ -n "$found" ] && error "found $uuid $nid on OST"
+
+ cleanup
+}
+run_test 91 "evict-by-nid support"
if ! combined_mgs_mds ; then
stop mgs