LU-3561 tests: improve sanity-hsm.sh to support remote agent

In the current sanity-hsm.sh, only one HSM copytool is started on
the local Lustre client node.

In order to test multiple remote HSM agents/copytools, this patch
introduces a new facet type named "AGT" and the following associated
variables:

- AGTCOUNT: number of agents
- AGTDEV{N}: target HSM mount point (root path of the backend)
- agt{N}_HOST: hostname of the agent agt{N}
- SINGLEAGT: facet of the single agent

By default, only single copytool is started on a remote client node.
If there was no remote client, then the copytool will be started on
the local client.

Signed-off-by: Jian Yu <jian.yu@intel.com>
Change-Id: I6e85d6948da894bd50148b0504347159ae23a79b
Reviewed-on: http://review.whamcloud.com/7110
Tested-by: Hudson
Reviewed-by: jacques-Charles Lafoucriere <jacques-charles.lafoucriere@cea.fr>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Jinshan Xiong <jinshan.xiong@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

diff --git a/lustre/tests/cfg/local.sh b/lustre/tests/cfg/local.sh
index 438c99f..13f06a3 100644 (file)
--- a/lustre/tests/cfg/local.sh
+++ b/lustre/tests/cfg/local.sh
@@ -125,3 +125,16 @@ MACHINEFILE_OPTION=${MACHINEFILE_OPTION:-"-machinefile"}
 # running the tests, or in some cases between the servers (e.g. lfsck.sh).
 # It needs to be a non-lustre filesystem that is available on all the nodes.
 SHARED_DIRECTORY=${SHARED_DIRECTORY:-$TMP}     # bug 17839 comment 65
 # running the tests, or in some cases between the servers (e.g. lfsck.sh).
 # It needs to be a non-lustre filesystem that is available on all the nodes.
 SHARED_DIRECTORY=${SHARED_DIRECTORY:-$TMP}     # bug 17839 comment 65

+
+#
+# In order to test multiple remote HSM agents, a new facet type named "AGT" and
+# the following associated variables are added:
+#
+# AGTCOUNT: number of agents
+# AGTDEV{N}: target HSM mount point (root path of the backend)
+# agt{N}_HOST: hostname of the agent agt{N}
+# SINGLEAGT: facet of the single agent
+#
+# Please refer to init_agt_vars() in sanity-hsm.sh for the default values of
+# these variables.
+#

diff --git a/lustre/tests/sanity-hsm.sh b/lustre/tests/sanity-hsm.sh
index 62164fb..7807177 100644 (file)
--- a/lustre/tests/sanity-hsm.sh
+++ b/lustre/tests/sanity-hsm.sh
@@ -30,7 +30,7 @@ MCREATE=${MCREATE:-mcreate}
 MOUNT_2=${MOUNT_2:-"yes"}
 FAIL_ON_ERROR=false
 
 MOUNT_2=${MOUNT_2:-"yes"}
 FAIL_ON_ERROR=false
 

-if [ $MDSCOUNT -ge 2 ]; then
+if [[ $MDSCOUNT -ge 2 ]]; then

        skip_env "Only run with single MDT for now" && exit
 fi
 
        skip_env "Only run with single MDT for now" && exit
 fi
 


@@ -41,77 +41,138 @@ if [[ $(lustre_version_code $SINGLEMDS) -lt $(version_code 2.4.53) ]]; then
 fi
 
 # $RUNAS_ID may get set incorrectly somewhere else
 fi
 
 # $RUNAS_ID may get set incorrectly somewhere else

-[ $UID -eq 0 -a $RUNAS_ID -eq 0 ] &&
-       error "\$RUNAS_ID set to 0, but \$UID is also 0!"
-
+if [[ $UID -eq 0 && $RUNAS_ID -eq 0 ]]; then
+       skip_env "\$RUNAS_ID set to 0, but \$UID is also 0!" && exit
+fi

 check_runas_id $RUNAS_ID $RUNAS_GID $RUNAS
 
 build_test_filter
 
 check_runas_id $RUNAS_ID $RUNAS_GID $RUNAS
 
 build_test_filter
 

-# the standard state when starting a test is
-# - no copytool
-# - MOUNT2 done
-# as some test changes the default, we need to re-make it
-cleanup() {
-       copytool_cleanup
-       if ! is_mounted $MOUNT2
-       then
-               mount_client $MOUNT2
+#
+# In order to test multiple remote HSM agents, a new facet type named "AGT" and
+# the following associated variables are added:
+#
+# AGTCOUNT: number of agents
+# AGTDEV{N}: target HSM mount point (root path of the backend)
+# agt{N}_HOST: hostname of the agent agt{N}
+# SINGLEAGT: facet of the single agent
+#
+# The number of agents is initialized as the number of remote client nodes.
+# By default, only single copytool is started on a remote client/agent. If there
+# was no remote client, then the copytool will be started on the local client.
+#
+init_agt_vars() {
+       local n
+       local agent
+
+       export AGTCOUNT=${AGTCOUNT:-$((CLIENTCOUNT - 1))}
+       [[ $AGTCOUNT -gt 0 ]] || AGTCOUNT=1
+
+       export SHARED_DIRECTORY=${SHARED_DIRECTORY:-$TMP}
+       if [[ $CLIENTCOUNT -gt 1 ]] &&
+               ! check_shared_dir $SHARED_DIRECTORY $CLIENTS; then
+               skip_env "SHARED_DIRECTORY should be accessible"\
+                        "on all client nodes"
+               exit 0

        fi
        fi

-       changelog_cleanup
+
+       for n in $(seq $AGTCOUNT); do
+               eval export AGTDEV$n=\$\{AGTDEV$n:-"$SHARED_DIRECTORY/arc$n"\}
+               agent=CLIENT$((n + 1))
+               if [[ -z "${!agent}" ]]; then
+                       [[ $CLIENTCOUNT -eq 1 ]] && agent=CLIENT1 ||
+                               agent=CLIENT2
+               fi
+               eval export agt${n}_HOST=\$\{agt${n}_HOST:-${!agent}\}
+       done
+
+       export SINGLEAGT=${SINGLEAGT:-agt1}
+
+       export HSMTOOL=${HSMTOOL:-"lhsmtool_posix"}
+       export HSMTOOL_VERBOSE=${HSMTOOL_VERBOSE:-""}
+       export HSMTOOL_BASE=$(basename "$HSMTOOL" | cut -f1 -d" ")
+       HSM_ARCHIVE=$(copytool_device $SINGLEAGT)
+       HSM_ARCHIVE_NUMBER=2
+
+       MDT_PARAM="mdt.$FSNAME-MDT0000"
+       HSM_PARAM="$MDT_PARAM.hsm"
+
+       # archive is purged at copytool setup
+       HSM_ARCHIVE_PURGE=true

 }
 
 }
 

-export HSMTOOL=${HSMTOOL:-"lhsmtool_posix"}
-export HSMTOOL_VERBOSE=${HSMTOOL_VERBOSE:-""}
-export HSMTOOL_BASE=$(basename "$HSMTOOL" | cut -f1 -d" ")
-HSM_ARCHIVE=${HSM_ARCHIVE:-$TMP/arc}
-HSM_ARCHIVE_NUMBER=2
+# Get the backend root path for the given agent facet.
+copytool_device() {
+       local facet=$1
+       local dev=AGTDEV$(facet_number $facet)

 
 

-MDT_PARAM="mdt.$FSNAME-MDT0000"
-HSM_PARAM="$MDT_PARAM.hsm"
+       echo -n ${!dev}
+}

 
 

-# archive is purged at copytool setup
-HSM_ARCHIVE_PURGE=true
+# Stop copytool and unregister an existing changelog user.
+cleanup() {
+       copytool_cleanup
+       changelog_cleanup
+}

 
 search_and_kill_copytool() {
 
 search_and_kill_copytool() {

-       echo "Killing existing copy tools"
-       killall -q $HSMTOOL_BASE || true
+       local agents=${1:-$(facet_active_host $SINGLEAGT)}
+
+       echo "Killing existing copytools on $agents"
+       do_nodesv $agents "killall -q $HSMTOOL_BASE" || true

 }
 
 copytool_setup() {
 }
 
 copytool_setup() {

-       if pkill -CONT -x $HSMTOOL_BASE; then
-               echo "Wakeup copytool"
-               return
+       local facet=${1:-$SINGLEAGT}
+       local lustre_mntpnt=${2:-$MOUNT}
+       local arc_id=$3
+       local hsm_root=$(copytool_device $facet)
+       local agent=$(facet_active_host $facet)
+
+       if [[ -z "$arc_id" ]] &&
+               do_facet $facet "pkill -CONT -x $HSMTOOL_BASE"; then
+                       echo "Wakeup copytool $facet on $agent"
+                       return 0

        fi
 
        if $HSM_ARCHIVE_PURGE; then
        fi
 
        if $HSM_ARCHIVE_PURGE; then

-               echo "Purging archive"
-               rm -rf $HSM_ARCHIVE/*
+               echo "Purging archive on $agent"
+               do_facet $facet "rm -rf $hsm_root/*"

        fi
 
        fi
 

-       echo "Starting copytool"
-       mkdir -p $HSM_ARCHIVE
+       echo "Starting copytool $facet on $agent"
+       do_facet $facet "mkdir -p $hsm_root" || error "mkdir '$hsm_root' failed"

        # bandwidth is limited to 1MB/s so the copy time is known and
        # independent of hardware
        # bandwidth is limited to 1MB/s so the copy time is known and
        # independent of hardware

-       local CMD="$HSMTOOL $HSMTOOL_VERBOSE --hsm-root $HSM_ARCHIVE"
-       CMD=$CMD" --daemon --bandwidth 1 $MOUNT"
-       [[ -z "$1" ]] || CMD+=" --archive $1"
-
-       echo "$CMD"
-       $CMD  &
+       local cmd="$HSMTOOL $HSMTOOL_VERBOSE --daemon --hsm-root $hsm_root"
+       [[ -z "$arc_id" ]] || cmd+=" --archive $arc_id"
+       cmd+=" --bandwidth 1 $lustre_mntpnt"
+
+       # Redirect the standard output and error to a log file which
+       # can be uploaded to Maloo.
+       local prefix=$TESTLOG_PREFIX
+       [[ -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"

        trap cleanup EXIT
 }
 
 copytool_cleanup() {
        trap - EXIT
        trap cleanup EXIT
 }
 
 copytool_cleanup() {
        trap - EXIT

-       pkill -INT -x $HSMTOOL_BASE || return 0
+       local agents=${1:-$(facet_active_host $SINGLEAGT)}
+
+       do_nodesv $agents "pkill -INT -x $HSMTOOL_BASE" || return 0

        sleep 1
        sleep 1

-       echo "Copytool is stopped"
+       echo "Copytool is stopped on $agents"

 }
 
 copytool_suspend() {
 }
 
 copytool_suspend() {

-       pkill -STOP -x $HSMTOOL_BASE || return 0
-       echo "Copytool is suspended"
+       local agents=${1:-$(facet_active_host $SINGLEAGT)}
+
+       do_nodesv $agents "pkill -STOP -x $HSMTOOL_BASE" || return 0
+       echo "Copytool is suspended on $agents"

 }
 
 copytool_remove_backend() {
 }
 
 copytool_remove_backend() {


@@ -122,8 +183,10 @@ copytool_remove_backend() {
 }
 
 import_file() {
 }
 
 import_file() {

-       $HSMTOOL --archive $HSM_ARCHIVE_NUMBER --hsm-root $HSM_ARCHIVE \
-               --import $1 $2 $MOUNT || error "import of $1 to $2 failed"
+       do_facet $SINGLEAGT \
+               "$HSMTOOL --archive $HSM_ARCHIVE_NUMBER --hsm-root $HSM_ARCHIVE\
+               --import $1 $2 $MOUNT" ||
+               error "import of $1 to $2 failed"

 }
 
 make_archive() {
 }
 
 make_archive() {


@@ -368,55 +431,55 @@ wait_result() {
        wait_update --verbose $(facet_active_host $facet) "$@"
 }
 
        wait_update --verbose $(facet_active_host $facet) "$@"
 }
 

-wait_request_state()
-{
+wait_request_state() {

        local fid=$1
        local request=$2
        local state=$3
        local fid=$1
        local request=$2
        local state=$3

-       wait_result $SINGLEMDS "$LCTL get_param -n $HSM_PARAM.agent_actions |\
-                               grep $fid | grep action=$request |\
-                               cut -f 13 -d ' ' | cut -f 2 -d =" $state 100 ||
+
+       local cmd="$LCTL get_param -n $HSM_PARAM.agent_actions"
+       cmd+=" | awk '/'$fid'.*action='$request'/ {print \\\$13}' | cut -f2 -d="
+
+       wait_result $SINGLEMDS "$cmd" $state 100 ||

                error "request on $fid is not $state"
 }
 
                error "request on $fid is not $state"
 }
 

-get_request_state()
-{
+get_request_state() {

        local fid=$1
        local request=$2
        local fid=$1
        local request=$2

-       do_facet $SINGLEMDS "$LCTL get_param -n $HSM_PARAM.agent_actions |\
-                               grep $fid | grep action=$request |\
-                               cut -f 13 -d ' ' | cut -f 2 -d ="
+
+       do_facet $SINGLEMDS "$LCTL get_param -n $HSM_PARAM.agent_actions |"\
+               "awk '/'$fid'.*action='$request'/ {print \\\$13}' | cut -f2 -d="

 }
 
 }
 

-get_request_count()
-{
+get_request_count() {

        local fid=$1
        local request=$2
        local fid=$1
        local request=$2

-       do_facet $SINGLEMDS "$LCTL get_param -n $HSM_PARAM.agent_actions |\
-                               grep $fid | grep action=$request | wc -l"
+
+       do_facet $SINGLEMDS "$LCTL get_param -n $HSM_PARAM.agent_actions |"\
+               "awk -vn=0 '/'$fid'.*action='$request'/ {n++}; END {print n}'"

 }
 
 }
 

-wait_all_done()
-{
+wait_all_done() {

        local timeout=$1
        local timeout=$1

-       wait_result $SINGLEMDS "$LCTL get_param -n $HSM_PARAM.agent_actions |\
-               egrep 'WAITING|STARTED' " "" $timeout ||
-       error "requests did not complete"
+
+       local cmd="$LCTL get_param -n $HSM_PARAM.agent_actions"
+       cmd+=" | egrep 'WAITING|STARTED'"
+
+       wait_result $SINGLEMDS "$cmd" "" $timeout ||
+               error "requests did not complete"

 }
 
 }
 

-wait_for_grace_delay()
-{
+wait_for_grace_delay() {

        local val=$(get_hsm_param grace_delay)
        sleep $val
 }
 
        local val=$(get_hsm_param grace_delay)
        sleep $val
 }
 

-my_uuid() {
-       $LCTL get_param -n llite.$FSNAME-*.uuid
-}
-

 MDT0=$($LCTL get_param -n mdc.*.mds_server_uuid |
        awk '{gsub(/_UUID/,""); print $1}' | head -1)
 
 MDT0=$($LCTL get_param -n mdc.*.mds_server_uuid |
        awk '{gsub(/_UUID/,""); print $1}' | head -1)
 

+# initiate variables
+init_agt_vars
+

 # cleanup from previous bad setup
 search_and_kill_copytool
 
 # cleanup from previous bad setup
 search_and_kill_copytool
 


@@ -583,9 +646,9 @@ test_9() {
        local f=$DIR/$tdir/$tfile
        local fid=$(copy_file /etc/passwd $f)
        # we do not use the default one to be sure
        local f=$DIR/$tdir/$tfile
        local fid=$(copy_file /etc/passwd $f)
        # we do not use the default one to be sure

-       local new_an=$((HSM_ARCHIVE_NUMBER+ 1))
+       local new_an=$((HSM_ARCHIVE_NUMBER + 1))

        copytool_cleanup
        copytool_cleanup

-       copytool_setup $new_an
+       copytool_setup $SINGLEAGT $MOUNT $new_an

        $LFS hsm_archive --archive $new_an $f
        wait_request_state $fid ARCHIVE SUCCEED
 
        $LFS hsm_archive --archive $new_an $f
        wait_request_state $fid ARCHIVE SUCCEED
 


@@ -595,6 +658,38 @@ test_9() {
 }
 run_test 9 "Use of explict archive number, with dedicated copytool"
 
 }
 run_test 9 "Use of explict archive number, with dedicated copytool"
 

+test_9a() {
+       [[ $CLIENTCOUNT -ge 3 ]] ||
+               { skip "Need three or more clients"; return 0; }
+
+       local n
+       local file
+       local fid
+
+       copytool_cleanup $(comma_list $(agts_nodes))
+
+       # start all of the copytools
+       for n in $(seq $AGTCOUNT); do
+               copytool_setup agt$n
+       done
+
+       trap "copytool_cleanup $(comma_list $(agts_nodes))" EXIT
+       # archive files
+       mkdir -p $DIR/$tdir
+       for n in $(seq $AGTCOUNT); do
+               file=$DIR/$tdir/$tfile.$n
+               fid=$(make_small $file)
+
+               $LFS hsm_archive $file || error "could not archive file $file"
+               wait_request_state $fid ARCHIVE SUCCEED
+               check_hsm_flags $file "0x00000001"
+       done
+
+       trap - EXIT
+       copytool_cleanup $(comma_list $(agts_nodes))
+}
+run_test 9a "Multiple remote agents"
+

 test_10a() {
        # test needs a running copytool
        copytool_setup
 test_10a() {
        # test needs a running copytool
        copytool_setup


@@ -976,9 +1071,10 @@ test_14() {
 
        # rebind the archive to the newly created file
        echo "rebind $fid to $fid2"
 
        # rebind the archive to the newly created file
        echo "rebind $fid to $fid2"

-       $HSMTOOL --archive $HSM_ARCHIVE_NUMBER --hsm-root="$HSM_ARCHIVE"\
-        --rebind $fid $fid2 $DIR ||
-               error "could not rebind file"
+
+       do_facet $SINGLEAGT \
+               "$HSMTOOL --archive $HSM_ARCHIVE_NUMBER --hsm-root $HSM_ARCHIVE\
+                --rebind $fid $fid2 $DIR" || error "could not rebind file"

 
        # restore file and compare md5sum
        local sum2=$(md5sum $f | awk '{print $1}')
 
        # restore file and compare md5sum
        local sum2=$(md5sum $f | awk '{print $1}')


@@ -997,7 +1093,7 @@ test_15() {
        mkdir -p $DIR/$tdir
        local f=$DIR/$tdir/$tfile
        local count=5
        mkdir -p $DIR/$tdir
        local f=$DIR/$tdir/$tfile
        local count=5

-       local tmpfile=$TMP/tmp.$$
+       local tmpfile=$SHARED_DIRECTORY/tmp.$$

 
        local fids=()
        local sums=()
 
        local fids=()
        local sums=()


@@ -1026,9 +1122,9 @@ test_15() {
        [[ $nl == $count ]] || error "$nl files in list, $count expected"
 
        echo "rebind list of files"
        [[ $nl == $count ]] || error "$nl files in list, $count expected"
 
        echo "rebind list of files"

-       $HSMTOOL --archive $HSM_ARCHIVE_NUMBER --hsm-root="$HSM_ARCHIVE"\
-        --rebind $tmpfile $DIR ||
-               error "could not rebind file list"
+       do_facet $SINGLEAGT \
+               "$HSMTOOL --archive $HSM_ARCHIVE_NUMBER --hsm-root $HSM_ARCHIVE\
+                --rebind $tmpfile $DIR" || error "could not rebind file list"

 
        # restore files and compare md5sum
        for i in $(seq 1 $count); do
 
        # restore files and compare md5sum
        for i in $(seq 1 $count); do


@@ -1425,8 +1521,7 @@ test_30b() {
 }
 run_test 30b "Restore at exec (release case)"
 
 }
 run_test 30b "Restore at exec (release case)"
 

-restore_and_check_size()
-{
+restore_and_check_size() {

        local f=$1
        local fid=$2
        local s=$(stat -c "%s" $f)
        local f=$1
        local fid=$2
        local s=$(stat -c "%s" $f)


@@ -1709,9 +1804,6 @@ test_52() {
        # test needs a running copytool
        copytool_setup
 
        # test needs a running copytool
        copytool_setup
 

-       # Test behave badly if 2 mount points are present
-       umount_client $MOUNT2
-

        mkdir -p $DIR/$tdir
        local f=$DIR/$tdir/$tfile
        local fid=$(copy_file /etc/motd $f 1)
        mkdir -p $DIR/$tdir
        local f=$DIR/$tdir/$tfile
        local fid=$(copy_file /etc/motd $f 1)


@@ -1731,9 +1823,6 @@ test_52() {
 
        check_hsm_flags $f "0x0000000b"
 
 
        check_hsm_flags $f "0x0000000b"
 

-       # Restore test environment
-       mount_client $MOUNT2
-

        copytool_cleanup
 }
 run_test 52 "Opened for write file on an evicted client should be set dirty"
        copytool_cleanup
 }
 run_test 52 "Opened for write file on an evicted client should be set dirty"


@@ -1742,9 +1831,6 @@ test_53() {
        # test needs a running copytool
        copytool_setup
 
        # test needs a running copytool
        copytool_setup
 

-       # Checks are wrong with 2 mount points
-       umount_client $MOUNT2
-

        mkdir -p $DIR/$tdir
        local f=$DIR/$tdir/$tfile
        local fid=$(copy_file /etc/motd $f 1)
        mkdir -p $DIR/$tdir
        local f=$DIR/$tdir/$tfile
        local fid=$(copy_file /etc/motd $f 1)


@@ -1765,8 +1851,6 @@ test_53() {
 
        check_hsm_flags $f "0x00000009"
 
 
        check_hsm_flags $f "0x00000009"
 

-       mount_client $MOUNT2
-

        copytool_cleanup
 }
 run_test 53 "Opened for read file on an evicted client should not be set dirty"
        copytool_cleanup
 }
 run_test 53 "Opened for read file on an evicted client should not be set dirty"


@@ -2098,13 +2182,11 @@ test_105() {
 run_test 105 "Restart of coordinator"
 
 test_106() {
 run_test 105 "Restart of coordinator"
 
 test_106() {

-       # Test behave badly if 2 mount points are present
-       umount_client $MOUNT2
-

        # test needs a running copytool
        copytool_setup
 
        # test needs a running copytool
        copytool_setup
 

-       local uuid=$(my_uuid)
+       local uuid=$(do_rpc_nodes $(facet_active_host $SINGLEAGT) \
+               get_client_uuid | cut -d' ' -f2)

        local agent=$(do_facet $SINGLEMDS $LCTL get_param -n $HSM_PARAM.agents |
                grep $uuid)
        copytool_cleanup
        local agent=$(do_facet $SINGLEMDS $LCTL get_param -n $HSM_PARAM.agents |
                grep $uuid)
        copytool_cleanup


@@ -2121,9 +2203,6 @@ test_106() {
        [[ ! -z "$agent" ]] ||
                error "My uuid $uuid not found in agent list after"\
                      " copytool restart"
        [[ ! -z "$agent" ]] ||
                error "My uuid $uuid not found in agent list after"\
                      " copytool restart"

-
-       # Restore test environment
-       mount_client $MOUNT2

 }
 run_test 106 "Copytool register/unregister"
 
 }
 run_test 106 "Copytool register/unregister"
 


@@ -2392,8 +2471,7 @@ test_221() {
        local target=0x7d
        [[ $flags == $target ]] || error "Changelog flag is $flags not $target"
 
        local target=0x7d
        [[ $flags == $target ]] || error "Changelog flag is $flags not $target"
 

-       changelog_cleanup
-       copytool_cleanup
+       cleanup

 }
 run_test 221 "Changelog for archive canceled"
 
 }
 run_test 221 "Changelog for archive canceled"
 


@@ -2417,8 +2495,7 @@ test_222a() {
        local target=0x80
        [[ $flags == $target ]] || error "Changelog flag is $flags not $target"
 
        local target=0x80
        [[ $flags == $target ]] || error "Changelog flag is $flags not $target"
 

-       changelog_cleanup
-       copytool_cleanup
+       cleanup

 }
 run_test 222a "Changelog for explicit restore"
 
 }
 run_test 222a "Changelog for explicit restore"
 


@@ -2444,8 +2521,7 @@ test_222b() {
        local target=0x80
        [[ $flags == $target ]] || error "Changelog flag is $flags not $target"
 
        local target=0x80
        [[ $flags == $target ]] || error "Changelog flag is $flags not $target"
 

-       changelog_cleanup
-       copytool_cleanup
+       cleanup

 }
 run_test 222b "Changelog for implicit restore"
 
 }
 run_test 222b "Changelog for implicit restore"
 


@@ -2475,8 +2551,7 @@ test_223a() {
        [[ $flags == $target ]] ||
                error "Changelog flag is $flags not $target"
 
        [[ $flags == $target ]] ||
                error "Changelog flag is $flags not $target"
 

-       changelog_cleanup
-       copytool_cleanup
+       cleanup

 }
 run_test 223a "Changelog for restore canceled (import case)"
 
 }
 run_test 223a "Changelog for restore canceled (import case)"
 


@@ -2505,8 +2580,7 @@ test_223b() {
        [[ $flags == $target ]] ||
                error "Changelog flag is $flags not $target"
 
        [[ $flags == $target ]] ||
                error "Changelog flag is $flags not $target"
 

-       changelog_cleanup
-       copytool_cleanup
+       cleanup

 }
 run_test 223b "Changelog for restore canceled (release case)"
 
 }
 run_test 223b "Changelog for restore canceled (release case)"
 


@@ -2532,8 +2606,7 @@ test_224() {
        [[ $flags == $target ]] ||
                error "Changelog flag is $flags not $target"
 
        [[ $flags == $target ]] ||
                error "Changelog flag is $flags not $target"
 

-       changelog_cleanup
-       copytool_cleanup
+       cleanup

 }
 run_test 224 "Changelog for remove"
 
 }
 run_test 224 "Changelog for remove"
 


@@ -2571,8 +2644,7 @@ test_225() {
        [[ $flags == $target ]] ||
                error "Changelog flag is $flags not $target"
 
        [[ $flags == $target ]] ||
                error "Changelog flag is $flags not $target"
 

-       changelog_cleanup
-       copytool_cleanup
+       cleanup

 }
 run_test 225 "Changelog for remove canceled"
 
 }
 run_test 225 "Changelog for remove canceled"
 


@@ -2612,8 +2684,7 @@ test_226() {
        [[ $flags == $target ]] ||
                error "Changelog flag is $flags not $target"
 
        [[ $flags == $target ]] ||
                error "Changelog flag is $flags not $target"
 

-       changelog_cleanup
-       copytool_cleanup
+       cleanup

 }
 run_test 226 "changelog for last rm/mv with exiting archive"
 
 }
 run_test 226 "changelog for last rm/mv with exiting archive"
 


@@ -2670,8 +2741,7 @@ test_227() {
        wait_request_state $fid ARCHIVE SUCCEED
        check_flags_changes $f $fid lost 3 1
 
        wait_request_state $fid ARCHIVE SUCCEED
        check_flags_changes $f $fid lost 3 1
 

-       changelog_cleanup
-       copytool_cleanup
+       cleanup

 }
 run_test 227 "changelog when explicit setting of HSM flags"
 
 }
 run_test 227 "changelog when explicit setting of HSM flags"
 

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index a643244..f1d54fb 100644 (file)
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -2334,14 +2334,32 @@ replay_barrier_nosync() {
        $LCTL mark "local REPLAY BARRIER on ${!svc}"
 }
 
        $LCTL mark "local REPLAY BARRIER on ${!svc}"
 }
 

+#
+# Get Lustre client uuid for a given Lustre mount point.
+#
+get_client_uuid() {
+       local mntpnt=${1:-$MOUNT}
+
+       local name=$($LFS getname $mntpnt | cut -d' ' -f1)
+       local uuid=$($LCTL get_param -n llite.$name.uuid)
+
+       echo -n $uuid
+}
+

 mds_evict_client() {
 mds_evict_client() {

-    UUID=`lctl get_param -n mdc.${mds1_svc}-mdc-*.uuid`
-    do_facet mds1 "lctl set_param -n mdt.${mds1_svc}.evict_client $UUID"
+       local mntpnt=${1:-$MOUNT}
+       local uuid=$(get_client_uuid $mntpnt)
+
+       do_facet $SINGLEMDS \
+               "$LCTL set_param -n mdt.${mds1_svc}.evict_client $uuid"

 }
 
 ost_evict_client() {
 }
 
 ost_evict_client() {

-    UUID=`lctl get_param -n devices| grep ${ost1_svc}-osc- | egrep -v 'MDT' | awk '{print $5}'`
-    do_facet ost1 "lctl set_param -n obdfilter.${ost1_svc}.evict_client $UUID"
+       local mntpnt=${1:-$MOUNT}
+       local uuid=$(get_client_uuid $mntpnt)
+
+       do_facet ost1 \
+               "$LCTL set_param -n obdfilter.${ost1_svc}.evict_client $uuid"

 }
 
 fail() {
 }
 
 fail() {


@@ -3726,10 +3744,11 @@ run_e2fsck() {
 # verify a directory is shared among nodes.
 check_shared_dir() {
        local dir=$1
 # verify a directory is shared among nodes.
 check_shared_dir() {
        local dir=$1

+       local list=${2:-$(comma_list $(nodes_list))}

 
        [ -z "$dir" ] && return 1
 
        [ -z "$dir" ] && return 1

-       do_rpc_nodes "$(comma_list $(nodes_list))" check_logdir $dir
-       check_write_access $dir || return 1
+       do_rpc_nodes "$list" check_logdir $dir
+       check_write_access $dir "$list" || return 1

        return 0
 }
 
        return 0
 }
 


@@ -3964,7 +3983,7 @@ get_facets () {
 
         case $type in
                 MGS ) list="$list $name";;
 
         case $type in
                 MGS ) list="$list $name";;

-            MDS|OST ) local count=${type}COUNT
+            MDS|OST|AGT ) local count=${type}COUNT

                        for ((i=1; i<=${!count}; i++)) do
                           list="$list ${name}$i"
                       done;;
                        for ((i=1; i<=${!count}; i++)) do
                           list="$list ${name}$i"
                       done;;


@@ -4755,6 +4774,11 @@ osts_nodes () {
        echo -n $(facets_nodes $(get_facets OST))
 }
 
        echo -n $(facets_nodes $(get_facets OST))
 }
 

+# Get all of the active AGT (HSM agent) nodes.
+agts_nodes () {
+       echo -n $(facets_nodes $(get_facets AGT))
+}
+

 # Get all of the client nodes and active server nodes.
 nodes_list () {
        local nodes=$HOSTNAME
 # Get all of the client nodes and active server nodes.
 nodes_list () {
        local nodes=$HOSTNAME


@@ -6003,19 +6027,20 @@ check_logdir() {
 }
 
 check_write_access() {
 }
 
 check_write_access() {

-    local dir=$1
-    local node
-    local file
+       local dir=$1
+       local list=${2:-$(comma_list $(nodes_list))}
+       local node
+       local file

 
 

-    for node in $(nodes_list); do
-        file=$dir/check_file.$(short_hostname $node)
-        if [[ ! -f "$file" ]]; then
-            # Logdir not accessible/writable from this node.
-            return 1
-        fi
-        rm -f $file || return 1
-    done
-    return 0
+       for node in ${list//,/ }; do
+               file=$dir/check_file.$(short_hostname $node)
+               if [[ ! -f "$file" ]]; then
+                       # Logdir not accessible/writable from this node.
+                       return 1
+               fi
+               rm -f $file || return 1
+       done
+       return 0

 }
 
 init_logging() {
 }
 
 init_logging() {