export PATH=$PWD/$SRCDIR:$SRCDIR:$PWD/$SRCDIR/utils:$PATH:/sbin:/usr/sbin
ONLY=${ONLY:-"$*"}
-# bug number for skipped test: 3815
+# bug number for skipped test: LU-3815
ALWAYS_EXCEPT="$SANITY_HSM_EXCEPT 34 35 36"
-# bug number for skipped test:4178 4176
-ALWAYS_EXCEPT="$ALWAYS_EXCEPT 200 221 223b 31a"
# UPDATE THE COMMENT ABOVE WITH BUG NUMBERS WHEN CHANGING ALWAYS_EXCEPT!
LUSTRE=${LUSTRE:-$(cd $(dirname $0)/..; echo $PWD)}
MULTIOP=${MULTIOP:-multiop}
OPENFILE=${OPENFILE:-openfile}
-MCREATE=${MCREATE:-mcreate}
+MMAP_CAT=${MMAP_CAT:-mmap_cat}
MOUNT_2=${MOUNT_2:-"yes"}
FAIL_ON_ERROR=false
build_test_filter
+# if there is no CLIENT1 defined, some tests can be ran on localhost
+CLIENT1=${CLIENT1:-$HOSTNAME}
+# if CLIENT2 doesn't exist then use CLIENT1 instead
+# All tests should use CLIENT2 with MOUNT2 only therefore it will work if
+# $CLIENT2 == CLIENT1
+# Exception is the test which need two separate nodes
+CLIENT2=${CLIENT2:-$CLIENT1}
+
#
# In order to test multiple remote HSM agents, a new facet type named "AGT" and
# the following associated variables are added:
exit 0
fi
+ # We used to put the HSM archive in $SHARED_DIRECTORY but that
+ # meant NFS issues could hose sanity-hsm sessions. So now we
+ # use $TMP instead.
for n in $(seq $AGTCOUNT); do
- eval export AGTDEV$n=\$\{AGTDEV$n:-"$SHARED_DIRECTORY/arc$n"\}
+ eval export AGTDEV$n=\$\{AGTDEV$n:-"$TMP/arc$n"\}
agent=CLIENT$((n + 1))
if [[ -z "${!agent}" ]]; then
[[ $CLIENTCOUNT -eq 1 ]] && agent=CLIENT1 ||
export HSMTOOL=${HSMTOOL:-"lhsmtool_posix"}
export HSMTOOL_VERBOSE=${HSMTOOL_VERBOSE:-""}
export HSMTOOL_UPDATE_INTERVAL=${HSMTOOL_UPDATE_INTERVAL:=""}
+ export HSMTOOL_EVENT_FIFO=${HSMTOOL_EVENT_FIFO:=""}
+ export HSMTOOL_TESTDIR
export HSMTOOL_BASE=$(basename "$HSMTOOL" | cut -f1 -d" ")
+ # $hsm_root/$HSMTMP Makes $hsm_root dir path less generic to ensure
+ # rm -rf $hsm_root/* is safe even if $hsm_root becomes unset to avoid
+ # deleting everything in filesystem, independent of any copytool.
+ export HSMTMP=${HSMTMP:-"shsm"}
+
HSM_ARCHIVE=$(copytool_device $SINGLEAGT)
+
+ [ -z "${HSM_ARCHIVE// /}" ] && error "HSM_ARCHIVE is empty!"
+ HSM_ARCHIVE=$HSM_ARCHIVE/$HSMTMP
+
HSM_ARCHIVE_NUMBER=2
# The test only support up to 10 MDTs
# archive is purged at copytool setup
HSM_ARCHIVE_PURGE=true
+
+ # Don't allow copytool error upon start/setup
+ HSMTOOL_NOERROR=false
}
# Get the backend root path for the given agent facet.
# Stop copytool and unregister an existing changelog user.
cleanup() {
+ copytool_monitor_cleanup
copytool_cleanup
changelog_cleanup
cdt_set_sanity_policy
local idx=$(($mdtno - 1))
MDT[$idx]=$($LCTL get_param -n \
mdc.$FSNAME-MDT000${idx}-mdc-*.mds_server_uuid |
- awk '{gsub(/_UUID/,""); print $1}' | head -1)
+ awk '{gsub(/_UUID/,""); print $1}' | head -n1)
done
}
search_copytools() {
- local agents=${1:-$(facet_active_host $SINGLEAGT)}
- do_nodesv $agents "pgrep -x $HSMTOOL_BASE"
+ local hosts=${1:-$(facet_active_host $SINGLEAGT)}
+ do_nodesv $hosts "pgrep -x $HSMTOOL_BASE"
}
-search_and_kill_copytool() {
- local agents=${1:-$(facet_active_host $SINGLEAGT)}
+kill_copytools() {
+ local hosts=${1:-$(facet_active_host $SINGLEAGT)}
+
+ echo "Killing existing copytools on $hosts"
+ do_nodesv $hosts "killall -q $HSMTOOL_BASE" || true
+}
+
+wait_copytools() {
+ local hosts=${1:-$(facet_active_host $SINGLEAGT)}
+ local wait_timeout=200
+ local wait_start=$SECONDS
+ local wait_end=$((wait_start + wait_timeout))
+
+ while ((SECONDS < wait_end)); do
+ sleep 2
+ if ! search_copytools $hosts; then
+ echo "copytools stopped in $((SECONDS - wait_start))s"
+ return 0
+ fi
+
+ echo "copytools still running on $hosts"
+ done
+
+ # try to dump Copytool's stack
+ do_nodesv $hosts "echo 1 >/proc/sys/kernel/sysrq ; " \
+ "echo t >/proc/sysrq-trigger"
+
+ echo "copytools failed to stop in ${wait_timeout}s"
+
+ return 1
+}
+
+copytool_monitor_setup() {
+ local facet=${1:-$SINGLEAGT}
+ local agent=$(facet_active_host $facet)
+
+ local cmd="mktemp --tmpdir=/tmp -d ${TESTSUITE}.${TESTNAME}.XXXX"
+ local test_dir=$(do_node $agent "$cmd") ||
+ error "Failed to create tempdir on $agent"
+ export HSMTOOL_MONITOR_DIR=$test_dir
+
+ # Create the fifo and a monitor (cat dies when copytool dies)
+ do_node $agent "mkfifo -m 0644 $test_dir/fifo" ||
+ error "failed to create copytool fifo on $agent"
+ cmd="cat $test_dir/fifo > $test_dir/events &"
+ cmd+=" echo \\\$! > $test_dir/monitor_pid"
+
+ if [[ $PDSH == *Rmrsh* ]]; then
+ # This is required for pdsh -Rmrsh and its handling of remote
+ # shells.
+ # Regular ssh and pdsh -Rssh work fine without this
+ # backgrounded subshell nonsense.
+ (do_node $agent "$cmd") &
+ export HSMTOOL_MONITOR_PDSH=$!
+
+ # Slightly racy, but just making a best-effort to catch obvious
+ # problems.
+ sleep 1
+ ps -p $HSMTOOL_MONITOR_PDSH > /dev/null ||
+ error "Failed to start copytool monitor on $agent"
+ else
+ do_node $agent "$cmd"
+ if [ $? != 0 ]; then
+ error "Failed to start copytool monitor on $agent"
+ fi
+ fi
+}
+
+copytool_monitor_cleanup() {
+ local facet=${1:-$SINGLEAGT}
+ local agent=$(facet_active_host $facet)
- echo "Killing existing copytools on $agents"
- do_nodesv $agents "killall -q $HSMTOOL_BASE" || true
+ if [ -n "$HSMTOOL_MONITOR_DIR" ]; then
+ # Should die when the copytool dies, but just in case.
+ local cmd="kill \\\$(cat $HSMTOOL_MONITOR_DIR/monitor_pid)"
+ cmd+=" 2>/dev/null || true"
+ do_node $agent "$cmd"
+ do_node $agent "rm -fr $HSMTOOL_MONITOR_DIR"
+ export HSMTOOL_MONITOR_DIR=
+ fi
+
+ # The pdsh should die on its own when the monitor dies. Just
+ # in case, though, try to clean up to avoid any cruft.
+ if [ -n "$HSMTOOL_MONITOR_PDSH" ]; then
+ kill $HSMTOOL_MONITOR_PDSH 2>/dev/null
+ export HSMTOOL_MONITOR_PDSH=
+ fi
}
copytool_setup() {
local facet=${1:-$SINGLEAGT}
- local lustre_mntpnt=${2:-$MOUNT}
+ # Use MOUNT2 by default if defined
+ local lustre_mntpnt=${2:-${MOUNT2:-$MOUNT}}
local arc_id=$3
local hsm_root=${4:-$(copytool_device $facet)}
+
+ [ -z "${hsm_root// /}" ] && error "copytool_setup: hsm_root empty!"
+
local agent=$(facet_active_host $facet)
if [[ -z "$arc_id" ]] &&
do_facet $facet "pkill -CONT -x $HSMTOOL_BASE"; then
- echo "Wakeup copytool $facet on $agent"
+ echo "Only wakeup running copytool $facet on $agent"
return 0
fi
if $HSM_ARCHIVE_PURGE; then
echo "Purging archive on $agent"
- do_facet $facet "rm -rf $hsm_root/*"
+ do_facet $facet "rm -rf $hsm_root/$HSMTMP/*"
fi
echo "Starting copytool $facet on $agent"
- do_facet $facet "mkdir -p $hsm_root" || error "mkdir '$hsm_root' failed"
+ do_facet $facet "mkdir -p $hsm_root/$HSMTMP/" ||
+ error "mkdir '$hsm_root/$HSMTMP' failed"
# bandwidth is limited to 1MB/s so the copy time is known and
# independent of hardware
- local cmd="$HSMTOOL $HSMTOOL_VERBOSE --daemon --hsm-root $hsm_root"
+ local cmd="$HSMTOOL $HSMTOOL_VERBOSE --daemon"
+ cmd+=" --hsm-root $hsm_root/$HSMTMP"
[[ -z "$arc_id" ]] || cmd+=" --archive $arc_id"
[[ -z "$HSMTOOL_UPDATE_INTERVAL" ]] ||
cmd+=" --update-interval $HSMTOOL_UPDATE_INTERVAL"
+ [[ -z "$HSMTOOL_EVENT_FIFO" ]] ||
+ cmd+=" --event-fifo $HSMTOOL_EVENT_FIFO"
cmd+=" --bandwidth 1 $lustre_mntpnt"
# Redirect the standard output and error to a log file which
[[ -z "$TESTNAME" ]] || prefix=$prefix.$TESTNAME
local copytool_log=$prefix.copytool${arc_id}_log.$agent.log
- do_facet $facet "$cmd < /dev/null > $copytool_log 2>&1" ||
- error "start copytool $facet on $agent failed"
+ do_facet $facet "$cmd < /dev/null > $copytool_log 2>&1"
+ if [[ $? != 0 ]]; then
+ [[ $HSMTOOL_NOERROR == true ]] ||
+ error "start copytool $facet on $agent failed"
+ echo "start copytool $facet on $agent failed"
+ fi
+
trap cleanup EXIT
}
+get_copytool_event_log() {
+ local facet=${1:-$SINGLEAGT}
+ local agent=$(facet_active_host $facet)
+
+ [ -z "$HSMTOOL_MONITOR_DIR" ] &&
+ error "Can't get event log: No monitor directory!"
+
+ do_node $agent "cat $HSMTOOL_MONITOR_DIR/events" ||
+ error "Could not collect event log from $agent"
+}
+
copytool_cleanup() {
trap - EXIT
- local agents=${1:-$(facet_active_host $SINGLEAGT)}
- local mdtno
- local idx
- local oldstate
- local mdt_hsmctrl
+ local agt_facet=$SINGLEAGT
+ local agt_hosts=${1:-$(facet_active_host $agt_facet)}
+ local hsm_root=$(copytool_device $agt_facet)
- do_nodesv $agents "pkill -INT -x $HSMTOOL_BASE" || return 0
- sleep 1
- echo "Copytool is stopped on $agents"
+ [ -z "${hsm_root// /}" ] && error "copytool_cleanup: hsm_root empty!"
- # clean all CDTs orphans requests from previous tests
- # that would otherwise need to timeout to clear.
- for mdtno in $(seq 1 $MDSCOUNT); do
- idx=$(($mdtno - 1))
- mdt_hsmctrl="mdt.$FSNAME-MDT000${idx}.hsm_control"
- oldstate=$(do_facet mds${mdtno} "$LCTL get_param -n " \
- "$mdt_hsmctrl")
- # skip already stop[ed,ing] CDTs
- echo $oldstate | grep stop && continue
-
- do_facet mds${mdtno} "$LCTL set_param $mdt_hsmctrl=shutdown"
- wait_result mds${mdtno} "$LCTL get_param -n $mdt_hsmctrl" \
- "stopped" 20 ||
- error "mds${mdtno} cdt state is not stopped"
- do_facet mds${mdtno} "$LCTL set_param $mdt_hsmctrl=$oldstate"
- wait_result mds${mdtno} "$LCTL get_param -n $mdt_hsmctrl" \
- "$oldstate" 20 ||
- error "mds${mdtno} cdt state is not $oldstate"
+ local i
+ local facet
+ local param
+ local -a state
+
+ kill_copytools $agt_hosts
+ wait_copytools $agt_hosts || error "copytools failed to stop"
+
+ # Clean all CDTs orphans requests from previous tests that
+ # would otherwise need to timeout to clear.
+ for ((i = 0; i < MDSCOUNT; i++)); do
+ facet=mds$((i + 1))
+ param=$(printf 'mdt.%s-MDT%04x.hsm_control' $FSNAME $i)
+ state[$i]=$(do_facet $facet "$LCTL get_param -n $param")
+
+ # Skip already stopping or stopped CDTs.
+ [[ "${state[$i]}" =~ ^stop ]] && continue
+
+ do_facet $facet "$LCTL set_param $param=shutdown"
done
+
+ for ((i = 0; i < MDSCOUNT; i++)); do
+ # Only check and restore CDTs that we stopped in the first loop.
+ [[ "${state[$i]}" =~ ^stop ]] && continue
+
+ facet=mds$((i + 1))
+ param=$(printf 'mdt.%s-MDT%04x.hsm_control' $FSNAME $i)
+
+ wait_result $facet "$LCTL get_param -n $param" stopped 20 ||
+ error "$facet CDT state is not stopped"
+
+ # Restore old CDT state.
+ do_facet $facet "$LCTL set_param $param=${state[$i]}"
+ done
+
+ for ((i = 0; i < MDSCOUNT; i++)); do
+ # Only check CDTs that we stopped in the first loop.
+ [[ "${state[$i]}" =~ ^stop ]] && continue
+
+ facet=mds$((i + 1))
+ param=$(printf 'mdt.%s-MDT%04x.hsm_control' $FSNAME $i)
+
+ # Check that the old CDT state was restored.
+ wait_result $facet "$LCTL get_param -n $param" "${state[$i]}" \
+ 20 || error "$facet CDT state is not '${state[$i]}'"
+ done
+
+ if do_facet $agt_facet "df $hsm_root" >/dev/null 2>&1 ; then
+ do_facet $agt_facet "rm -rf $hsm_root/$HSMTMP/*"
+ fi
}
copytool_suspend() {
local file=$HSM_ARCHIVE/$1
do_facet $SINGLEAGT mkdir -p $(dirname $file)
do_facet $SINGLEAGT dd if=/dev/urandom of=$file count=32 bs=1000000 ||
- error "cannot create $file"
+ file_creation_failure dd $file $?
}
copy2archive() {
cdt_set_mount_state() {
mdts_set_param "-P" hsm_control "$1"
+ # set_param -P is asynchronous operation and could race with set_param.
+ # In such case configs could be retrieved and applied at mgc after
+ # set_param -P completion. Sleep here to avoid race with set_param.
+ # We need at least 20 seconds. 10 for mgc_requeue_thread to wake up
+ # MGC_TIMEOUT_MIN_SECONDS + MGC_TIMEOUT_RAND_CENTISEC(5 + 5)
+ # and 10 seconds to retrieve config from server.
+ sleep 20
}
cdt_check_state() {
path2fid() {
$LFS path2fid $1 | tr -d '[]'
+ return ${PIPESTATUS[0]}
}
get_hsm_flags() {
local f=$1
local u=$2
+ local st
if [[ $u == "user" ]]; then
- local st=$($RUNAS $LFS hsm_state $f)
+ st=$($RUNAS $LFS hsm_state $f)
else
- local st=$($LFS hsm_state $f)
u=root
+ st=$($LFS hsm_state $f)
fi
[[ $? == 0 ]] || error "$LFS hsm_state $f failed (run as $u)"
get_hsm_archive_id() {
local f=$1
- local st=$($LFS hsm_state $f)
+ local st
+ st=$($LFS hsm_state $f)
[[ $? == 0 ]] || error "$LFS hsm_state $f failed"
local ar=$(echo $st | grep "archive_id" | cut -f5 -d" " |
[[ $st == $fl ]] || error "hsm flags on $f are $st != $fl"
}
+file_creation_failure() {
+ local cmd=$1
+ local f=$2
+ local err=$3
+
+ df $MOUNT $MOUNT2 >&2
+ error "cannot create $f with $cmd, status=$err"
+}
+
copy_file() {
local f=
f=${f/$DIR/$DIR2}
fi
rm -f $f
- cp $1 $f || error "cannot copy $1 to $f"
+ cp $1 $f || file_creation_failure cp $f $?
+
path2fid $f || error "cannot get fid on $f"
}
make_small() {
local file2=${1/$DIR/$DIR2}
dd if=/dev/urandom of=$file2 count=2 bs=1M conv=fsync ||
- error "cannot create $file2"
+ file_creation_failure dd $file2 $?
+
path2fid $1 || error "cannot get fid on $1"
}
make_small_sync() {
dd if=/dev/urandom of=$1 count=1 bs=1M conv=sync ||
- error "cannot create $1"
+ file_creation_failure dd $1 $?
path2fid $1 || error "cannot get fid on $1"
}
[ $ratio -gt 50 ] && find $MOUNT -size +10M -exec rm -f {} \;
}
+check_enough_free_space() {
+ local nb=$1
+ local unit=$2
+ local need=$((nb * unit /1024))
+ local free=$(df -kP $MOUNT | tail -1 | awk '{print $4}')
+ (( $need >= $free )) && return 1
+ return 0
+}
+
make_large_for_striping() {
local file2=${1/$DIR/$DIR2}
- local sz=$($LCTL get_param -n lov.*-clilov-*.stripesize | head -1)
+ local sz=$($LCTL get_param -n lov.*-clilov-*.stripesize | head -n1)
cleanup_large_files
+ check_enough_free_space 5 $sz
+ [ $? != 0 ] && return $?
+
dd if=/dev/urandom of=$file2 count=5 bs=$sz conv=fsync ||
- error "cannot create $file2"
+ file_creation_failure dd $file2 $?
+
path2fid $1 || error "cannot get fid on $1"
}
cleanup_large_files
+ check_enough_free_space 39 1000000
+ [ $? != 0 ] && return $?
+
# big file is large enough, so copy time is > 30s
# so copytool make 1 progress
# size is not a multiple of 1M to avoid stripe
# aligment
dd if=/dev/urandom of=$file2 count=39 bs=1000000 conv=fsync ||
- error "cannot create $file2"
+ file_creation_failure dd $file2 $?
+
path2fid $1 || error "cannot get fid on $1"
}
cleanup_large_files
+ check_enough_free_space 33 1048576
+ [ $? != 0 ] && return $?
+
# big file is large enough, so copy time is > 30s
# so copytool make 1 progress
# size is a multiple of 1M to have stripe
# aligment
dd if=/dev/urandom of=$file2 count=33 bs=1M conv=fsync ||
- error "cannot create $file2"
+ file_creation_failure dd $file2 $?
path2fid $1 || error "cannot get fid on $1"
}
cleanup_large_files
+ check_enough_free_space 103 1048576
+ [ $? != 0 ] && return $?
+
# Copy timeout is 100s. 105MB => 105s
dd if=/dev/urandom of=$file2 count=103 bs=1M conv=fsync ||
- error "cannot create $file2"
+ file_creation_failure dd $file2 $?
path2fid $1 || error "cannot get fid on $1"
}
local cmd="$LCTL get_param -n ${MDT_PREFIX}${mdtidx}.hsm.actions"
cmd+=" | awk '/'$fid'.*action='$request'/ {print \\\$13}' | cut -f2 -d="
- wait_result $mds "$cmd" $state 100 ||
+ wait_result $mds "$cmd" $state 200 ||
error "request on $fid is not $state on $mds"
}
wait_all_done() {
local timeout=$1
+ local fid=$2
local cmd="$LCTL get_param -n $HSM_PARAM.actions"
+ [[ -n $fid ]] && cmd+=" | grep '$fid'"
cmd+=" | egrep 'WAITING|STARTED'"
wait_result $SINGLEMDS "$cmd" "" $timeout ||
sleep $val
}
-# populate MDT device array
-get_mdt_devices
+parse_json_event() {
+ local raw_event=$1
+
+ # python2.6 in EL6 includes an internal json module
+ local json_parser='import json; import fileinput;'
+ json_parser+=' print "\n".join(["local %s=\"%s\"" % tuple for tuple in '
+ json_parser+='json.loads([line for line in '
+ json_parser+='fileinput.input()][0]).items()])'
+
+ echo $raw_event | python -c "$json_parser"
+}
# initiate variables
init_agt_vars
+# populate MDT device array
+get_mdt_devices
+
# cleanup from previous bad setup
-search_and_kill_copytool
+kill_copytools
# for recovery tests, coordinator needs to be started at mount
# so force it
}
run_test 1 "lfs hsm flags root/non-root access"
+test_1a() {
+ mkdir -p $DIR/$tdir
+ local f=$DIR/$tdir/$tfile
+ local fid=$(make_small $f)
+
+ $LFS hsm_archive $f || error "could not archive file"
+ wait_request_state $fid ARCHIVE SUCCEED
+
+ # Release and check states
+ $LFS hsm_release $f || error "could not release file"
+ echo -n "Verifying released state: "
+ check_hsm_flags $f "0x0000000d"
+
+ $MMAP_CAT $f > /dev/null || error "failed mmap & cat release file"
+}
+run_test 1a "mmap & cat a HSM released file"
+
test_2() {
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
error "user could not change hsm flags"
dd if=/etc/passwd of=$f.append bs=1 count=3\
conv=notrunc oflag=append status=noxfer ||
- error "could not append to test file"
+ file_creation_failure dd $f.append $?
check_hsm_flags $f.append "0x00000003"
# Modify a file sets it dirty
error "user could not change hsm flags"
dd if=/dev/zero of=$f.modify bs=1 count=3\
conv=notrunc status=noxfer ||
- error "could not modify test file"
+ file_creation_failure dd $f.modify $?
check_hsm_flags $f.modify "0x00000003"
# Open O_TRUNC sets dirty
copytool_cleanup
}
-run_test 9 "Use of explict archive number, with dedicated copytool"
+run_test 9 "Use of explicit archive number, with dedicated copytool"
test_9a() {
needclients 3 || return 0
}
run_test 10d "Archive a file on the default archive id"
-test_11() {
+test_11a() {
mkdir -p $DIR/$tdir
copy2archive /etc/hosts $tdir/$tfile
local f=$DIR/$tdir/$tfile
local AFILE=$(do_facet $SINGLEAGT ls $HSM_ARCHIVE'/*/*/*/*/*/*/'$fid) ||
error "fid $fid not in archive $HSM_ARCHIVE"
}
-run_test 11 "Import a file"
+run_test 11a "Import a file"
+
+test_11b() {
+ # test needs a running copytool
+ copytool_setup
+
+ mkdir -p $DIR/$tdir
+ local f=$DIR/$tdir/$tfile
+ local fid=$(copy_file /etc/hosts $f)
+ $LFS hsm_archive -a $HSM_ARCHIVE_NUMBER $f ||
+ error "hsm_archive failed"
+ wait_request_state $fid ARCHIVE SUCCEED
+
+ local FILE_HASH=$(md5sum $f)
+ rm -f $f
+
+ import_file $fid $f
+
+ echo "$FILE_HASH" | md5sum -c
+
+ [[ $? -eq 0 ]] || error "Restored file differs"
+
+ copytool_cleanup
+}
+run_test 11b "Import a deleted file using its FID"
test_12a() {
# test needs a running copytool
local f=$DIR/$tdir/$tfile
import_file $tdir/$tfile $f
- local f=$DIR2/$tdir/$tfile
+ local f2=$DIR2/$tdir/$tfile
echo "Verifying released state: "
- check_hsm_flags $f "0x0000000d"
+ check_hsm_flags $f2 "0x0000000d"
- local fid=$(path2fid $f)
- $LFS hsm_restore $f
+ local fid=$(path2fid $f2)
+ $LFS hsm_restore $f2
wait_request_state $fid RESTORE SUCCEED
echo "Verifying file state: "
- check_hsm_flags $f "0x00000009"
+ check_hsm_flags $f2 "0x00000009"
do_facet $SINGLEAGT diff -q $HSM_ARCHIVE/$tdir/$tfile $f
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
$LFS setstripe -c 2 $f
- local fid=$(make_large_for_striping $f)
+ local fid
+ fid=$(make_large_for_striping $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
local FILE_CRC=$(md5sum $f)
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
}
run_test 12o "Layout-swap failure during Restore leaves file released"
+test_12p() {
+ # test needs a running copytool
+ copytool_setup
+
+ mkdir $DIR/$tdir
+ local f=$DIR/$tdir/$tfile
+ local fid=$(copy_file /etc/hosts $f)
+
+ $LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
+ wait_request_state $fid ARCHIVE SUCCEED
+ do_facet $SINGLEAGT cat $f > /dev/null || error "cannot cat $f"
+ $LFS hsm_release $f || error "cannot release $f"
+ do_facet $SINGLEAGT cat $f > /dev/null || error "cannot cat $f"
+ $LFS hsm_release $f || error "cannot release $f"
+ do_facet $SINGLEAGT cat $f > /dev/null || error "cannot cat $f"
+
+ copytool_cleanup
+}
+run_test 12p "implicit restore of a file on copytool mount point"
+
+cleanup_test_12q() {
+ trap 0
+ zconf_umount $(facet_host $SINGLEAGT) $MOUNT3 ||
+ error "cannot umount $MOUNT3 on $SINGLEAGT"
+}
+
+test_12q() {
+ [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.58) ] &&
+ skip "need MDS version at least 2.7.58" && return 0
+
+ zconf_mount $(facet_host $SINGLEAGT) $MOUNT3 ||
+ error "cannot mount $MOUNT3 on $SINGLEAGT"
+
+ trap cleanup_test_12q EXIT
+
+ # test needs a running copytool
+ copytool_setup $SINGLEAGT $MOUNT3
+
+ mkdir $DIR/$tdir
+ local f=$DIR/$tdir/$tfile
+ local f2=$DIR2/$tdir/$tfile
+ local fid=$(make_small $f)
+ local orig_size=$(stat -c "%s" $f)
+
+ $LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
+ wait_request_state $fid ARCHIVE SUCCEED
+
+ $LFS hsm_release $f || error "could not release file"
+ check_hsm_flags $f "0x0000000d"
+
+ kill_copytools
+ wait_copytools || error "copytool failed to stop"
+
+ cat $f > /dev/null &
+
+ # wait a bit to allow implicit restore request to be handled.
+ # if not, next stat would also block on layout-lock.
+ sleep 5
+
+ local size=$(stat -c "%s" $f2)
+ [ $size -eq $orig_size ] ||
+ error "$f2: wrong size after archive: $size != $orig_size"
+
+ HSM_ARCHIVE_PURGE=false copytool_setup $SINGLEAGT /mnt/lustre3
+
+ wait
+
+ size=$(stat -c "%s" $f)
+ [ $size -eq $orig_size ] ||
+ error "$f: wrong size after restore: $size != $orig_size"
+
+ size=$(stat -c "%s" $f2)
+ [ $size -eq $orig_size ] ||
+ error "$f2: wrong size after restore: $size != $orig_size"
+
+ :>$f
+
+ size=$(stat -c "%s" $f)
+ [ $size -eq 0 ] ||
+ error "$f: wrong size after overwrite: $size != 0"
+
+ size=$(stat -c "%s" $f2)
+ [ $size -eq 0 ] ||
+ error "$f2: wrong size after overwrite: $size != 0"
+
+ copytool_cleanup
+ zconf_umount $(facet_host $SINGLEAGT) $MOUNT3 ||
+ error "cannot umount $MOUNT3 on $SINGLEAGT"
+}
+run_test 12q "file attributes are refreshed after restore"
+
test_13() {
# test needs a running copytool
copytool_setup
[ $mtime0 -eq $mtime1 ] ||
error "restore changed mtime from $mtime0 to $mtime1"
- [ $ctime0 -le $ctime1 ] ||
+ [ $ctime0 -eq $ctime1 ] ||
error "restore changed ctime from $ctime0 to $ctime1"
copytool_cleanup
[ $mtime0 -eq $mtime1 ] ||
error "remount changed mtime from $mtime0 to $mtime1"
- [ $ctime0 -le $ctime1 ] ||
+ [ $ctime0 -eq $ctime1 ] ||
error "remount changed ctime from $ctime0 to $ctime1"
}
run_test 24a "Archive, release, and restore does not change a/mtime (i/o)"
copytool_setup
mkdir -p $DIR/$tdir
- # Check that root can do HSM actions on a ordinary user's file.
+ # Check that root can do HSM actions on a regular user's file.
rm -f $file
fid=$(make_small $file)
sum0=$(md5sum $file)
}
run_test 24d "check that read-only mounts are respected"
+test_24e() {
+ copytool_setup
+
+ mkdir -p $DIR/$tdir
+
+ local f=$DIR/$tdir/$tfile
+ local fid
+
+ fid=$(make_small $f) || error "cannot create $f"
+ $LFS hsm_archive $f || error "cannot archive $f"
+ wait_request_state $fid ARCHIVE SUCCEED
+ $LFS hsm_release $f || error "cannot release $f"
+ while ! $LFS hsm_state $f | grep released; do
+ sleep 1
+ done
+
+ tar -cf $TMP/$tfile.tar $DIR/$tdir || error "cannot tar $DIR/$tdir"
+
+ copytool_cleanup
+}
+run_test 24e "tar succeeds on HSM released files" # LU-6213
+
+test_24f() {
+
+ # test needs a running copytool
+ copytool_setup
+
+ mkdir -p $DIR/$tdir/d1
+ local f=$DIR/$tdir/$tfile
+ local fid=$(copy_file /etc/hosts $f)
+ sum0=$(md5sum $f)
+ echo $sum0
+ $LFS hsm_archive -a $HSM_ARCHIVE_NUMBER $f ||
+ error "hsm_archive failed"
+ wait_request_state $fid ARCHIVE SUCCEED
+ $LFS hsm_release $f || error "cannot release $f"
+ tar --xattrs -cvf $f.tar -C $DIR/$tdir $tfile
+ rm -f $f
+ sync
+ tar --xattrs -xvf $f.tar -C $DIR/$tdir ||
+ error "Can not recover the tar contents"
+ sum1=$(md5sum $f)
+ echo "Sum0 = $sum0, sum1 = $sum1"
+ [ "$sum0" == "$sum1" ] || error "md5sum mismatch for '$tfile'"
+
+ copytool_cleanup
+}
+run_test 24f "root can archive, release, and restore tar files"
+
test_25a() {
# test needs a running copytool
copytool_setup
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE SUCCEED
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE SUCCEED
$LFS hsm_release $f
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE SUCCEED
}
run_test 28 "Concurrent archive/file remove"
+test_29a() {
+ # Tests --mntpath and --archive options
+
+ local archive_id=7
+ copytool_setup $SINGLEAGT $MOUNT $archive_id
+
+ # Bad archive number
+ $LFS hsm_remove -m $MOUNT -a 33 0x857765760:0x8:0x2 2>&1 |
+ grep "Invalid argument" ||
+ error "unexpected hsm_remove failure (1)"
+
+ # mntpath is present but file is given
+ $LFS hsm_remove --mntpath $MOUNT --archive 30 /qwerty/uyt 2>&1 |
+ grep "hsm: '/qwerty/uyt' is not a valid FID" ||
+ error "unexpected hsm_remove failure (2)"
+
+ copytool_cleanup
+}
+run_test 29a "Tests --mntpath and --archive options"
+
+test_29b() {
+ # test needs a running copytool
+ copytool_setup
+
+ mkdir -p $DIR/$tdir
+ local f=$DIR/$tdir/$tfile
+ local fid=$(make_small $f)
+
+ $LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
+ wait_request_state $fid ARCHIVE SUCCEED
+
+ rm -f $f
+
+ $LFS hsm_remove -m $MOUNT -a $HSM_ARCHIVE_NUMBER $fid
+ wait_request_state $fid REMOVE SUCCEED
+
+ copytool_cleanup
+}
+run_test 29b "Archive/delete/remove by FID from the archive."
+
+test_29c() {
+ # test needs a running copytool
+ copytool_setup
+
+ mkdir -p $DIR/$tdir
+ local fid1=$(make_small $DIR/$tdir/$tfile-1)
+ local fid2=$(make_small $DIR/$tdir/$tfile-2)
+ local fid3=$(make_small $DIR/$tdir/$tfile-3)
+
+ $LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $DIR/$tdir/$tfile-[1-3]
+ wait_request_state $fid1 ARCHIVE SUCCEED
+ wait_request_state $fid2 ARCHIVE SUCCEED
+ wait_request_state $fid3 ARCHIVE SUCCEED
+
+ rm -f $DIR/$tdir/$tfile-[1-3]
+
+ echo $fid1 > $DIR/$tdir/list
+ echo $fid2 >> $DIR/$tdir/list
+ echo $fid3 >> $DIR/$tdir/list
+
+ $LFS hsm_remove -m $MOUNT -a $HSM_ARCHIVE_NUMBER \
+ --filelist $DIR/$tdir/list
+ wait_request_state $fid1 REMOVE SUCCEED
+ wait_request_state $fid2 REMOVE SUCCEED
+ wait_request_state $fid3 REMOVE SUCCEED
+
+ copytool_cleanup
+}
+run_test 29c "Archive/delete/remove by FID, using a file list."
+
test_30a() {
# restore at exec cannot work on agent node (because of Linux kernel
# protection of executables)
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/SLEEP
+ local slp_sum1=$(md5sum /bin/sleep)
local fid=$(copy_file /bin/sleep $f)
chmod 755 $f
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait $pid
[[ $? == 0 ]] || error "Execution failed during run"
cmp /bin/sleep $f
- [[ $? == 0 ]] || error "Binary overwritten during exec"
+ if [[ $? != 0 ]]; then
+ local slp_sum2=$(md5sum /bin/sleep)
+ # in case sleep file is modified during the test
+ [[ $slp_sum1 == $slp_sum2 ]] &&
+ error "Binary overwritten during exec"
+ fi
# cleanup
# remove no try action mode
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE SUCCEED
$LFS hsm_release $f
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress_aligned $f)
+ local fid
+ fid=$(make_large_for_progress_aligned $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE SUCCEED
$LFS hsm_release $f
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE SUCCEED
$LFS hsm_release $f
+ # to be sure wait_all_done will not be mislead by previous tests
+ # and ops.
+ cdt_purge
+ wait_for_grace_delay
+ # Also raise grace_delay significantly so the Canceled
+ # Restore action will stay enough long avail.
+ local old_grace=$(get_hsm_param grace_delay)
+ set_hsm_param grace_delay 100
+
md5sum $f >/dev/null &
local pid=$!
wait_request_state $fid RESTORE STARTED
$LFS hsm_cancel $f
- wait_request_state $fid RESTORE CANCELED
- wait_request_state $fid CANCEL SUCCEED
+ # instead of waiting+checking both Restore and Cancel ops
+ # sequentially, wait for both to be finished and then check
+ # each results.
+ wait_all_done 100 $fid
+ local rstate=$(get_request_state $fid RESTORE)
+ local cstate=$(get_request_state $fid CANCEL)
+
+ # restore orig grace_delay.
+ set_hsm_param grace_delay $old_grace
+
+ if [[ "$rstate" == "CANCELED" ]] ; then
+ [[ "$cstate" == "SUCCEED" ]] ||
+ error "Restore state is CANCELED and Cancel state " \
+ "is not SUCCEED but $cstate"
+ echo "Restore state is CANCELED, Cancel state is SUCCEED"
+ elif [[ "$rstate" == "SUCCEED" ]] ; then
+ [[ "$cstate" == "FAILED" ]] ||
+ error "Restore state is SUCCEED and Cancel state " \
+ "is not FAILED but $cstate"
+ echo "Restore state is SUCCEED, Cancel state is FAILED"
+ else
+ error "Restore state is $rstate and Cancel state is $cstate"
+ fi
[ -z $killed ] ||
error "Cannot kill process waiting for restore ($killed)"
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE SUCCEED
$LFS hsm_release $f
local f=$DIR/$tdir/$tfile
local f1=$DIR/$tdir/$tfile-1
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
local fid1=$(copy_file /etc/passwd $f1)
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE SUCCEED
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE SUCCEED
$LFS hsm_release $f
}
run_test 36 "Move file during restore"
+test_37() {
+ # LU-5683: check that an archived dirty file can be rearchived.
+ copytool_cleanup
+ copytool_setup $SINGLEAGT $MOUNT2
+
+ mkdir -p $DIR/$tdir
+ local f=$DIR/$tdir/$tfile
+ local fid
+
+ fid=$(make_small $f) || error "cannot create small file"
+
+ $LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
+ wait_request_state $fid ARCHIVE SUCCEED
+ $LFS hsm_release $f || error "cannot release $f"
+
+ # Dirty file.
+ dd if=/dev/urandom of=$f bs=1M count=1 || error "cannot dirty file"
+
+ $LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
+ wait_request_state $fid ARCHIVE SUCCEED
+
+ copytool_cleanup
+}
+run_test 37 "re-archive a dirty file"
+
multi_archive() {
local prefix=$1
local count=$2
done
# force copytool to use a local/temp archive dir to ensure best
# performance vs remote/NFS mounts used in auto-tests
- if df --local $HSM_ARCHIVE >/dev/null 2>&1 ; then
+ if do_facet $SINGLEAGT "df --local $HSM_ARCHIVE" >/dev/null 2>&1 ; then
copytool_setup
else
copytool_setup $SINGLEAGT $MOUNT $HSM_ARCHIVE_NUMBER $TMP/$tdir
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_small $f)
+ local fid=$(make_large_for_progress $f)
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f ||
error "could not archive file"
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_small $f)
+ local fid=$(make_large_for_progress $f)
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f ||
error "could not archive file"
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f ||
error "could not archive file"
}
run_test 58 "Truncate a released file will trigger restore"
+test_59() {
+ local fid
+ local server_version=$(lustre_version_code $SINGLEMDS)
+ [[ $server_version -lt $(version_code 2.7.63) ]] &&
+ skip "Need MDS version at least 2.7.63" && return
+
+ copytool_setup
+ $MCREATE $DIR/$tfile || error "mcreate failed"
+ $TRUNCATE $DIR/$tfile 42 || error "truncate failed"
+ $LFS hsm_archive $DIR/$tfile || error "archive request failed"
+ fid=$(path2fid $DIR/$tfile)
+ wait_request_state $fid ARCHIVE SUCCEED
+ $LFS hsm_release $DIR/$tfile || error "release failed"
+ copytool_cleanup
+}
+run_test 59 "Release stripeless file with non-zero size"
+
test_60() {
# This test validates the fix for LU-4512. Ensure that the -u
- # option changes the progress reporting interval from the default
- # (30 seconds) to the user-specified interval.
+ # option changes the progress reporting interval from the
+ # default (30 seconds) to the user-specified interval.
local interval=5
- local progress_timeout=$((interval * 2))
+ local progress_timeout=$((interval * 4))
# test needs a new running copytool
copytool_cleanup
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
+ local mdtidx=0
+ local mdt=${MDT_PREFIX}${mdtidx}
+ local mds=mds$((mdtidx + 1))
+
+ # Wait for copytool to register
+ wait_update_facet $mds \
+ "$LCTL get_param -n ${mdt}.hsm.agents | grep -o ^uuid" \
+ uuid 100 || error "coyptool failed to register with $mdt"
local start_at=$(date +%s)
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f ||
error "could not archive file"
- local mdtidx=0
- local mdt=${MDT_PREFIX}${mdtidx}
- local mds=mds$((mdtidx + 1))
+ local agent=$(facet_active_host $SINGLEAGT)
+ local prefix=$TESTLOG_PREFIX
+ [[ -z "$TESTNAME" ]] || prefix=$prefix.$TESTNAME
+ local copytool_log=$prefix.copytool_log.$agent.log
+
+
+ wait_update $agent \
+ "grep -o start.copy $copytool_log" "start copy" 100 ||
+ error "copytool failed to start"
local cmd="$LCTL get_param -n ${mdt}.hsm.active_requests"
cmd+=" | awk '/'$fid'.*action=ARCHIVE/ {print \\\$12}' | cut -f2 -d="
local elapsed=$((finish_at - start_at))
# Ensure that the progress update occurred within the expected window.
- if [ $elapsed -gt $progress_timeout ]; then
- error "Expected progress update within $progress_timeout seconds"
- elif [ $elapsed -lt $interval ]; then
+ if [ $elapsed -lt $((interval - 1)) ]; then
error "Expected progress update after at least $interval seconds"
fi
}
run_test 60 "Changing progress update interval from default"
+test_70() {
+ # test needs a new running copytool
+ copytool_cleanup
+ copytool_monitor_setup
+ HSMTOOL_EVENT_FIFO=$HSMTOOL_MONITOR_DIR/fifo copytool_setup
+
+ # Just start and stop the copytool to generate events.
+ cdt_clear_no_retry
+
+ # Wait for the copytool to register.
+ wait_update --verbose $(facet_active_host mds1) \
+ "$LCTL get_param -n ${MDT_PREFIX}0.hsm.agents | grep -o ^uuid" \
+ uuid 100 ||
+ error "copytool failed to register with MDT0000"
+
+ copytool_cleanup
+
+ local REGISTER_EVENT
+ local UNREGISTER_EVENT
+ while read event; do
+ local parsed=$(parse_json_event "$event")
+ if [ -z "$parsed" ]; then
+ error "Copytool sent malformed event: $event"
+ fi
+ eval $parsed
+
+ if [ $event_type == "REGISTER" ]; then
+ REGISTER_EVENT=$event
+ elif [ $event_type == "UNREGISTER" ]; then
+ UNREGISTER_EVENT=$event
+ fi
+ done < <(echo $"$(get_copytool_event_log)")
+
+ if [ -z "$REGISTER_EVENT" ]; then
+ error "Copytool failed to send register event to FIFO"
+ fi
+
+ if [ -z "$UNREGISTER_EVENT" ]; then
+ error "Copytool failed to send unregister event to FIFO"
+ fi
+
+ copytool_monitor_cleanup
+ echo "Register/Unregister events look OK."
+}
+run_test 70 "Copytool logs JSON register/unregister events to FIFO"
+
+test_71() {
+ # Bump progress interval for livelier events.
+ local interval=5
+
+ # test needs a new running copytool
+ copytool_cleanup
+ copytool_monitor_setup
+ HSMTOOL_UPDATE_INTERVAL=$interval \
+ HSMTOOL_EVENT_FIFO=$HSMTOOL_MONITOR_DIR/fifo copytool_setup
+
+ mkdir -p $DIR/$tdir
+ local f=$DIR/$tdir/$tfile
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
+ $LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f ||
+ error "could not archive file"
+ wait_request_state $fid ARCHIVE SUCCEED
+
+ local expected_fields="event_time data_fid source_fid"
+ expected_fields+=" total_bytes current_bytes"
+
+ local START_EVENT
+ local FINISH_EVENT
+ while read event; do
+ # Make sure we're not getting anything from previous events.
+ for field in $expected_fields; do
+ unset $field
+ done
+
+ local parsed=$(parse_json_event "$event")
+ if [ -z "$parsed" ]; then
+ error "Copytool sent malformed event: $event"
+ fi
+ eval $parsed
+
+ if [ $event_type == "ARCHIVE_START" ]; then
+ START_EVENT=$event
+ continue
+ elif [ $event_type == "ARCHIVE_FINISH" ]; then
+ FINISH_EVENT=$event
+ continue
+ elif [ $event_type != "ARCHIVE_RUNNING" ]; then
+ continue
+ fi
+
+ # Do some simple checking of the progress update events.
+ for expected_field in $expected_fields; do
+ if [ -z ${!expected_field+x} ]; then
+ error "Missing $expected_field field in event"
+ fi
+ done
+
+ if [ $total_bytes -eq 0 ]; then
+ error "Expected total_bytes to be > 0"
+ fi
+
+ # These should be identical throughout an archive
+ # operation.
+ if [ $source_fid != $data_fid ]; then
+ error "Expected source_fid to equal data_fid"
+ fi
+ done < <(echo $"$(get_copytool_event_log)")
+
+ if [ -z "$START_EVENT" ]; then
+ error "Copytool failed to send archive start event to FIFO"
+ fi
+
+ if [ -z "$FINISH_EVENT" ]; then
+ error "Copytool failed to send archive finish event to FIFO"
+ fi
+
+ echo "Archive events look OK."
+
+ cdt_clear_no_retry
+ copytool_cleanup
+ copytool_monitor_cleanup
+}
+run_test 71 "Copytool logs JSON archive events to FIFO"
+
+test_72() {
+ # Bump progress interval for livelier events.
+ local interval=5
+
+ # test needs a new running copytool
+ copytool_cleanup
+ copytool_monitor_setup
+ HSMTOOL_UPDATE_INTERVAL=$interval \
+ HSMTOOL_EVENT_FIFO=$HSMTOOL_MONITOR_DIR/fifo copytool_setup
+ local test_file=$HSMTOOL_MONITOR_DIR/file
+
+ local cmd="dd if=/dev/urandom of=$test_file count=16 bs=1000000 "
+ cmd+="conv=fsync"
+ do_facet $SINGLEAGT "$cmd" ||
+ error "cannot create $test_file on $SINGLEAGT"
+ copy2archive $test_file $tdir/$tfile
+
+ mkdir -p $DIR/$tdir
+ local f=$DIR/$tdir/$tfile
+ import_file $tdir/$tfile $f
+ f=$DIR2/$tdir/$tfile
+ echo "Verifying released state: "
+ check_hsm_flags $f "0x0000000d"
+
+ local fid=$(path2fid $f)
+ $LFS hsm_restore $f
+ wait_request_state $fid RESTORE SUCCEED
+
+ local expected_fields="event_time data_fid source_fid"
+ expected_fields+=" total_bytes current_bytes"
+
+ local START_EVENT
+ local FINISH_EVENT
+ while read event; do
+ # Make sure we're not getting anything from previous events.
+ for field in $expected_fields; do
+ unset $field
+ done
+
+ local parsed=$(parse_json_event "$event")
+ if [ -z "$parsed" ]; then
+ error "Copytool sent malformed event: $event"
+ fi
+ eval $parsed
+
+ if [ $event_type == "RESTORE_START" ]; then
+ START_EVENT=$event
+ if [ $source_fid != $data_fid ]; then
+ error "source_fid should == data_fid at start"
+ fi
+ continue
+ elif [ $event_type == "RESTORE_FINISH" ]; then
+ FINISH_EVENT=$event
+ if [ $source_fid != $data_fid ]; then
+ error "source_fid should == data_fid at finish"
+ fi
+ continue
+ elif [ $event_type != "RESTORE_RUNNING" ]; then
+ continue
+ fi
+
+ # Do some simple checking of the progress update events.
+ for expected_field in $expected_fields; do
+ if [ -z ${!expected_field+x} ]; then
+ error "Missing $expected_field field in event"
+ fi
+ done
+
+ if [ $total_bytes -eq 0 ]; then
+ error "Expected total_bytes to be > 0"
+ fi
+
+ # When a restore starts out, the data fid is the same as the
+ # source fid. After the restore has gotten going, we learn
+ # the new data fid. Once the restore has finished, the source
+ # fid is set to the new data fid.
+ #
+ # We test this because some monitoring software may depend on
+ # this behavior. If it changes, then the consumers of these
+ # events may need to be modified.
+ if [ $source_fid == $data_fid ]; then
+ error "source_fid should != data_fid during restore"
+ fi
+ done < <(echo $"$(get_copytool_event_log)")
+
+ if [ -z "$START_EVENT" ]; then
+ error "Copytool failed to send restore start event to FIFO"
+ fi
+
+ if [ -z "$FINISH_EVENT" ]; then
+ error "Copytool failed to send restore finish event to FIFO"
+ fi
+
+ echo "Restore events look OK."
+
+ cdt_clear_no_retry
+ copytool_cleanup
+ copytool_monitor_cleanup
+
+ rm -rf $test_dir
+}
+run_test 72 "Copytool logs JSON restore events to FIFO"
+
test_90() {
file_count=51 # Max number of files constrained by LNET message size
mkdir $DIR/$tdir || error "mkdir $DIR/$tdir failed"
fid=$(copy_file /etc/hosts $f.$i)
echo $f.$i >> $FILELIST
done
- copytool_setup
+ # force copytool to use a local/temp archive dir to ensure best
+ # performance vs remote/NFS mounts used in auto-tests
+ if do_facet $SINGLEAGT "df --local $HSM_ARCHIVE" >/dev/null 2>&1 ; then
+ copytool_setup
+ else
+ local dai=$(get_hsm_param default_archive_id)
+ copytool_setup $SINGLEAGT $MOUNT $dai $TMP/$tdir
+ fi
# to be sure wait_all_done will not be mislead by previous tests
cdt_purge
wait_for_grace_delay
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
# if cdt is on, it can serve too quickly the request
cdt_disable
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER --data $DATA $f
done
}
-test_106() {
- local uuid=$(do_rpc_nodes $(facet_active_host $SINGLEAGT) \
- get_client_uuid $MOUNT | cut -d' ' -f2)
+get_agent_uuid() {
+ local agent=${1:-$(facet_active_host $SINGLEAGT)}
+ # Lustre mount-point is mandatory and last parameter on
+ # copytool cmd-line.
+ local mntpnt=$(do_rpc_nodes $agent pgrep -fl $HSMTOOL_BASE |
+ grep -v pgrep | awk '{print $NF}')
+ [ -n "$mntpnt" ] || error "Found no Agent or with no mount-point "\
+ "parameter"
+ do_rpc_nodes $agent get_client_uuid $mntpnt | cut -d' ' -f2
+}
+
+test_106() {
+ # test needs a running copytool
copytool_setup
+
+ local uuid=$(get_agent_uuid $(facet_active_host $SINGLEAGT))
+
check_agent_registered $uuid
search_copytools || error "No copytool found"
check_agent_unregistered $uuid
copytool_setup
+ uuid=$(get_agent_uuid $(facet_active_host $SINGLEAGT))
check_agent_registered $uuid
copytool_cleanup
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_cancel $f)
+ local fid
+ fid=$(make_large_for_cancel $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
# test with cdt on is made in test_221
cdt_disable
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
+ # wait archive to register at CDT
+ wait_request_state $fid ARCHIVE WAITING
$LFS hsm_cancel $f
cdt_enable
wait_request_state $fid ARCHIVE CANCELED
# test with cdt on is made in test_222
cdt_disable
$LFS hsm_restore $f
+ # wait restore to register at CDT
+ wait_request_state $fid RESTORE WAITING
$LFS hsm_cancel $f
cdt_enable
wait_request_state $fid RESTORE CANCELED
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
+
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE SUCCEED
cdt_disable
$LFS hsm_remove $f
+ # wait remove to register at CDT
+ wait_request_state $fid REMOVE WAITING
$LFS hsm_cancel $f
cdt_enable
wait_request_state $fid REMOVE CANCELED
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_cancel $f)
+ local fid
+ fid=$(make_large_for_cancel $f)
+ [ $? != 0 ] && skip "not enough free space" && return
changelog_setup
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
changelog_setup
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_progress $f)
+ local fid
+ fid=$(make_large_for_progress $f)
+ [ $? != 0 ] && skip "not enough free space" && return
changelog_setup
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
mkdir -p $DIR/$tdir
local f=$DIR/$tdir/$tfile
- local fid=$(make_large_for_cancel $f)
+ local fid
+ fid=$(make_large_for_cancel $f)
+ [ $? != 0 ] && skip "not enough free space" && return
cdt_disable
# to have a short test
set_hsm_param loop_period 2
cdt_enable
+ # clear locks to avoid extra delay caused by flush/cancel
+ # and thus prevent early copytool death to timeout.
+ cancel_lru_locks osc
+
$LFS hsm_archive --archive $HSM_ARCHIVE_NUMBER $f
wait_request_state $fid ARCHIVE STARTED
sleep 5
copytool_cleanup
# deactivate all mdc on agent1
- mdc_change_state $SINGLEAGT "MDT000." "deactivate"
+ mdc_change_state $SINGLEAGT "$FSNAME-MDT000." "deactivate"
- copytool_setup $SINGLEAGT
+ HSMTOOL_NOERROR=true copytool_setup $SINGLEAGT
check_agent_unregistered "uuid" # match any agent
search_copytools $agent && error "Copytool start should have failed"
# reactivate MDCs
- mdc_change_state $SINGLEAGT "MDT000." "activate"
+ mdc_change_state $SINGLEAGT "$FSNAME-MDT000." "activate"
}
run_test 402 "Copytool start fails if all MDTs are inactive"
copytool_cleanup
local agent=$(facet_active_host $SINGLEAGT)
- local uuid=$(do_rpc_nodes $agent get_client_uuid | cut -d' ' -f2)
# deactivate all mdc for MDT0001
- mdc_change_state $SINGLEAGT "MDT0001" "deactivate"
+ mdc_change_state $SINGLEAGT "$FSNAME-MDT0001" "deactivate"
copytool_setup
+ local uuid=$(get_agent_uuid $agent)
# check the agent is registered on MDT0000, and not on MDT0001
check_agent_registered_by_mdt $uuid 0
check_agent_unregistered_by_mdt $uuid 1
search_copytools $agent || error "No running copytools on $agent"
# reactivate all mdc for MDT0001
- mdc_change_state $SINGLEAGT "MDT0001" "activate"
+ mdc_change_state $SINGLEAGT "$FSNAME-MDT0001" "activate"
# make sure the copytool is now registered to all MDTs
check_agent_registered $uuid
local fid1=$(make_small $dir_mdt0/$tfile)
# deactivate all mdc for MDT0001
- mdc_change_state $SINGLEAGT "MDT0001" "deactivate"
+ mdc_change_state $SINGLEAGT "$FSNAME-MDT0001" "deactivate"
# send an HSM request for files in MDT0000
$LFS hsm_archive $dir_mdt0/$tfile || error "lfs hsm_archive"
echo "archive successful on mdt0"
# reactivate all mdc for MDT0001
- mdc_change_state $SINGLEAGT "MDT0001" "activate"
+ mdc_change_state $SINGLEAGT "$FSNAME-MDT0001" "activate"
copytool_cleanup
# clean test files and directories
}
run_test 404 "Inactive MDT does not block requests for active MDTs"
+test_405() {
+ [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return
+
+ copytool_setup
+
+ mkdir -p $DIR/$tdir
+
+ local striped_dir=$DIR/$tdir/striped_dir
+
+ # create striped dir on all of MDTs
+ $LFS mkdir -i 0 -c $MDSCOUNT $striped_dir || error "lfs mkdir"
+
+ local fid1=$(make_small_sync $striped_dir/${tfile}_0)
+ local fid2=$(make_small_sync $striped_dir/${tfile}_1)
+ local fid3=$(make_small_sync $striped_dir/${tfile}_2)
+ local fid4=$(make_small_sync $striped_dir/${tfile}_3)
+
+ local idx1=$($LFS getstripe -M $striped_dir/${tfile}_0)
+ local idx2=$($LFS getstripe -M $striped_dir/${tfile}_1)
+ local idx3=$($LFS getstripe -M $striped_dir/${tfile}_2)
+ local idx4=$($LFS getstripe -M $striped_dir/${tfile}_3)
+
+ # check that compound requests are shunt to the rights MDTs
+ $LFS hsm_archive $striped_dir/${tfile}_0 $striped_dir/${tfile}_1 \
+ $striped_dir/${tfile}_2 $striped_dir/${tfile}_3 ||
+ error "lfs hsm_archive"
+
+ wait_request_state $fid1 ARCHIVE SUCCEED $idx1 &&
+ echo "archive successful on $fid1"
+ wait_request_state $fid2 ARCHIVE SUCCEED $idx2 &&
+ echo "archive successful on $fid2"
+ wait_request_state $fid3 ARCHIVE SUCCEED $idx3 &&
+ echo "archive successful on $fid3"
+ wait_request_state $fid4 ARCHIVE SUCCEED $idx4 &&
+ echo "archive successful on $fid4"
+
+ $LFS hsm_release $striped_dir/${tfile}_0 || error "lfs hsm_release 1"
+ $LFS hsm_release $striped_dir/${tfile}_1 || error "lfs hsm_release 2"
+ $LFS hsm_release $striped_dir/${tfile}_2 || error "lfs hsm_release 3"
+ $LFS hsm_release $striped_dir/${tfile}_3 || error "lfs hsm_release 4"
+
+ cat $striped_dir/${tfile}_0 > /dev/null || error "cat ${tfile}_0 failed"
+ cat $striped_dir/${tfile}_1 > /dev/null || error "cat ${tfile}_1 failed"
+ cat $striped_dir/${tfile}_2 > /dev/null || error "cat ${tfile}_2 failed"
+ cat $striped_dir/${tfile}_3 > /dev/null || error "cat ${tfile}_3 failed"
+
+ copytool_cleanup
+}
+run_test 405 "archive and release under striped directory"
+
+test_406() {
+ [ $MDSCOUNT -lt 2 ] && skip "needs >= 2 MDTs" && return 0
+
+ [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.7.64) ] &&
+ skip "need MDS version at least 2.7.64" && return 0
+
+ local fid
+ local mdt_index
+
+ copytool_setup
+ mkdir -p $DIR/$tdir
+ fid=$(make_small $DIR/$tdir/$tfile)
+ echo "old fid $fid"
+
+ $LFS hsm_archive $DIR/$tdir/$tfile
+ wait_request_state "$fid" ARCHIVE SUCCEED
+ $LFS hsm_release $DIR/$tdir/$tfile
+
+ # Should migrate $tdir but not $tfile.
+ $LFS mv -M1 $DIR/$tdir &&
+ error "migrating HSM an archived file should fail"
+
+ $LFS hsm_restore $DIR/$tdir/$tfile
+ wait_request_state "$fid" RESTORE SUCCEED
+
+ $LFS hsm_remove $DIR/$tdir/$tfile
+ wait_request_state "$fid" REMOVE SUCCEED
+
+ cat $DIR/$tdir/$tfile > /dev/null ||
+ error "cannot read $DIR/$tdir/$tfile"
+
+ $LFS mv -M1 $DIR/$tdir ||
+ error "cannot complete migration after HSM remove"
+
+ mdt_index=$($LFS getstripe -M $DIR/$tdir)
+ if ((mdt_index != 1)); then
+ error "expected MDT index 1, got $mdt_index"
+ fi
+
+ # Refresh fid after migration.
+ fid=$(path2fid $DIR/$tdir/$tfile)
+ echo "new fid $fid"
+
+ $LFS hsm_archive $DIR/$tdir/$tfile
+ wait_request_state "$fid" ARCHIVE SUCCEED 1
+
+ lctl set_param debug=+trace
+ $LFS hsm_release $DIR/$tdir/$tfile ||
+ error "cannot release $DIR/$tdir/$tfile"
+
+ $LFS hsm_restore $DIR/$tdir/$tfile
+ wait_request_state "$fid" RESTORE SUCCEED 1
+
+ cat $DIR/$tdir/$tfile > /dev/null ||
+ error "cannot read $DIR/$tdir/$tfile"
+
+ copytool_cleanup
+}
+run_test 406 "attempting to migrate HSM archived files is safe"
+
+test_500()
+{
+ [ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.6.92) ] &&
+ skip "HSM migrate is not supported" && return
+
+ # Stop the existing copytool
+ copytool_cleanup
+
+ test_mkdir -p $DIR/$tdir
+ llapi_hsm_test -d $DIR/$tdir || error "One llapi HSM test failed"
+}
+run_test 500 "various LLAPI HSM tests"
+
copytool_cleanup
complete $SECONDS