1 # Hey Emacs, this is a -*- shell-script -*- !!! :-)
3 .
"${TEST_SCRIPTS_DIR}/common.sh"
5 ######################################################################
9 if [ -n "$CTDB_TEST_REMOTE_DIR" ] ; then
10 CTDB_TEST_WRAPPER
="${CTDB_TEST_REMOTE_DIR}/test_wrap"
12 _d
=$
(cd ${TEST_SCRIPTS_DIR}; echo $PWD)
13 CTDB_TEST_WRAPPER
="$_d/test_wrap"
15 export CTDB_TEST_WRAPPER
17 # If $VALGRIND is set then use it whenever ctdb is called, but only if
18 # $CTDB is not already set.
19 [ -n "$CTDB" ] ||
export CTDB
="${VALGRIND}${VALGRIND:+ }ctdb"
22 PATH
="${TEST_SCRIPTS_DIR}:${PATH}"
24 ######################################################################
32 [ $
(($testfailures+0)) -eq 0 -a $status -ne 0 ] && testfailures
=$status
33 status
=$
(($testfailures+0))
35 # Avoid making a test fail from this point onwards. The test is
39 echo "*** TEST COMPLETED (RC=$status) AT $(date '+%F %T'), CLEANING UP..."
41 eval "$ctdb_test_exit_hook" || true
42 unset ctdb_test_exit_hook
44 if $ctdb_test_restart_scheduled ||
! cluster_is_healthy
; then
48 # This could be made unconditional but then we might get
49 # duplication from the recovery in restart_ctdb. We want to
50 # leave the recovery in restart_ctdb so that future tests that
51 # might do a manual restart mid-test will benefit.
52 echo "Forcing a recovery..."
53 onnode
0 $CTDB recover
59 ctdb_test_exit_hook_add
()
61 ctdb_test_exit_hook
="${ctdb_test_exit_hook}${ctdb_test_exit_hook:+ ; }$*"
66 scriptname
=$
(basename "$0")
68 ctdb_test_restart_scheduled
=false
70 trap "ctdb_test_exit" 0
73 ########################################
76 try_command_on_node
()
78 local nodespec
="$1" ; shift
83 while [ "${nodespec#-}" != "$nodespec" ] ; do
84 if [ "$nodespec" = "-v" ] ; then
87 onnode_opts
="${onnode_opts}${onnode_opts:+ }${nodespec}"
94 out
=$
(onnode
-q $onnode_opts "$nodespec" "$cmd" 2>&1) ||
{
96 echo "Failed to execute \"$cmd\" on node(s) \"$nodespec\""
102 echo "Output of \"$cmd\":"
107 sanity_check_output
()
110 local regexp
="$2" # Should be anchored as necessary.
115 local num_lines
=$
(echo "$output" |
wc -l)
116 echo "There are $num_lines lines of output"
117 if [ $num_lines -lt $min_lines ] ; then
118 echo "BAD: that's less than the required number (${min_lines})"
123 local unexpected
# local doesn't pass through status of command on RHS.
124 unexpected
=$
(echo "$output" |
egrep -v "$regexp") || status
=$?
126 # Note that this is reversed.
127 if [ $status -eq 0 ] ; then
128 echo "BAD: unexpected lines in output:"
129 echo "$unexpected" |
cat -A
132 echo "Output lines look OK"
140 local ips
="$1" # list of "ip node" lines
142 echo "Sanity checking IPs..."
146 while read x ipp
; do
147 [ "$ipp" = "-1" ] && break
148 if [ -n "$prev" -a "$ipp" != "$prev" ] ; then
155 echo "BAD: a node was -1 or IPs are only assigned to one node:"
157 echo "Are you running an old version of CTDB?"
161 # This returns a list of "ip node" lines in $out
165 try_command_on_node
$node \
166 "$CTDB ip -X | awk -F'|' 'NR > 1 { print \$2, \$3 }'"
169 _select_test_node_and_ips
()
171 try_command_on_node any \
172 "$CTDB ip -X all | awk -F'|' 'NR > 1 { print \$2, \$3 }'"
174 test_node
="" # this matches no PNN
177 while read ip pnn
; do
178 if [ -z "$test_node" -a "$pnn" != "-1" ] ; then
181 if [ "$pnn" = "$test_node" ] ; then
182 test_node_ips
="${test_node_ips}${test_node_ips:+ }${ip}"
184 done <<<"$out" # bashism to avoid problem setting variable in pipeline.
186 echo "Selected node ${test_node} with IPs: ${test_node_ips}."
187 test_ip
="${test_node_ips%% *}"
190 *:*) test_prefix
="${test_ip}/128" ;;
191 *) test_prefix
="${test_ip}/32" ;;
194 [ -n "$test_node" ] ||
return 1
197 select_test_node_and_ips
()
200 while ! _select_test_node_and_ips
; do
201 echo "Unable to find a test node with IPs assigned"
202 if [ $timeout -le 0 ] ; then
203 echo "BAD: Too many attempts"
207 timeout
=$
(($timeout - 1))
214 get_test_ip_mask_and_iface
()
217 try_command_on_node
$test_node "$CTDB ip -v -X | awk -F'|' -v ip=$test_ip '\$2 == ip { print \$4 }'"
220 if [ -z "$TEST_LOCAL_DAEMONS" ] ; then
222 try_command_on_node
$test_node ip addr show to
$test_ip
229 echo "$test_ip/$mask is on $iface"
234 try_command_on_node
-q all
"$CTDB pnn | sed -e 's@PNN:@@'"
238 # The subtlety is that "ctdb delip" will fail if the IP address isn't
239 # configured on a node...
240 delete_ip_from_all_nodes
()
248 for _pnn
in $all_pnns ; do
249 all_ips_on_node
$_pnn
250 while read _i _n
; do
251 if [ "$_ip" = "$_i" ] ; then
252 _nodes
="${_nodes}${_nodes:+,}${_pnn}"
254 done <<<"$out" # bashism
257 try_command_on_node
-pq "$_nodes" "$CTDB delip $_ip"
260 #######################################
262 # Wait until either timeout expires or command succeeds. The command
263 # will be tried once per second, unless timeout has format T/I, where
264 # I is the recheck interval.
267 local timeout
="$1" ; shift # "$@" is the command...
272 interval
="${timeout#*/}"
273 timeout
="${timeout%/*}"
277 if [ "$1" = "!" ] ; then
282 echo -n "<${timeout}|"
284 while [ $t -gt 0 ] ; do
287 if { ! $negate && [ $rc -eq 0 ] ; } || \
288 { $negate && [ $rc -ne 0 ] ; } ; then
289 echo "|$(($timeout - $t))|"
294 for i
in $
(seq 1 $interval) ; do
297 t
=$
(($t - $interval))
309 for i
in $
(seq 1 $1) ; do
316 _cluster_is_healthy
()
318 $CTDB nodestatus all
>/dev
/null
321 _cluster_is_recovered
()
323 node_has_status
0 recovered
328 _cluster_is_healthy
&& _cluster_is_recovered
331 cluster_is_healthy
()
333 if onnode
0 $CTDB_TEST_WRAPPER _cluster_is_healthy
; then
334 echo "Cluster is HEALTHY"
335 if ! onnode
0 $CTDB_TEST_WRAPPER _cluster_is_recovered
; then
336 echo "WARNING: cluster in recovery mode!"
340 echo "Cluster is UNHEALTHY"
341 if ! ${ctdb_test_restart_scheduled:-false} ; then
342 echo "DEBUG AT $(date '+%F %T'):"
344 for i
in "onnode -q 0 $CTDB status" "onnode -q 0 onnode all $CTDB scriptstatus" ; do
355 local timeout
="${1:-120}"
357 echo "Waiting for cluster to become ready..."
359 wait_until
$timeout onnode
-q any
$CTDB_TEST_WRAPPER _cluster_is_ready
362 # This function is becoming nicely overloaded. Soon it will collapse! :-)
368 local bits fpat mpat rpat
370 (unhealthy
) bits
="?|?|?|1|*" ;;
371 (healthy
) bits
="?|?|?|0|*" ;;
372 (disconnected
) bits
="1|*" ;;
373 (connected
) bits
="0|*" ;;
374 (banned
) bits
="?|1|*" ;;
375 (unbanned
) bits
="?|0|*" ;;
376 (disabled
) bits
="?|?|1|*" ;;
377 (enabled
) bits
="?|?|0|*" ;;
378 (stopped
) bits
="?|?|?|?|1|*" ;;
379 (notstopped
) bits
="?|?|?|?|0|*" ;;
380 (frozen
) fpat
='^[[:space:]]+frozen[[:space:]]+1$' ;;
381 (unfrozen
) fpat
='^[[:space:]]+frozen[[:space:]]+0$' ;;
382 (monon
) mpat
='^Monitoring mode:ACTIVE \(0\)$' ;;
383 (monoff
) mpat
='^Monitoring mode:DISABLED \(1\)$' ;;
384 (recovered
) rpat
='^Recovery mode:RECOVERY \(1\)$' ;;
386 echo "node_has_status: unknown status \"$status\""
390 if [ -n "$bits" ] ; then
393 out
=$
($CTDB -X status
2>&1) ||
return 1
398 # This needs to be done in 2 steps to avoid false matches.
399 local line_bits
="${line#|${pnn}|*|}"
400 [ "$line_bits" = "$line" ] && continue
401 [ "${line_bits#${bits}}" != "$line_bits" ] && return 0
404 } <<<"$out" # Yay bash!
405 elif [ -n "$fpat" ] ; then
406 $CTDB statistics
-n "$pnn" |
egrep -q "$fpat"
407 elif [ -n "$mpat" ] ; then
408 $CTDB getmonmode
-n "$pnn" |
egrep -q "$mpat"
409 elif [ -n "$rpat" ] ; then
410 ! $CTDB status
-n "$pnn" |
egrep -q "$rpat"
412 echo 'node_has_status: unknown mode, neither $bits nor $fpat is set'
417 wait_until_node_has_status
()
421 local timeout
="${3:-30}"
422 local proxy_pnn
="${4:-any}"
424 echo "Waiting until node $pnn has status \"$status\"..."
426 if ! wait_until
$timeout onnode
$proxy_pnn $CTDB_TEST_WRAPPER node_has_status
"$pnn" "$status" ; then
427 for i
in "onnode -q any $CTDB status" "onnode -q any onnode all $CTDB scriptstatus" ; do
437 # Useful for superficially testing IP failover.
438 # IPs must be on the given node.
439 # If the first argument is '!' then the IPs must not be on the given node.
443 if [ "$1" = "!" ] ; then
444 negating
=true
; shift
446 local node
="$1" ; shift
451 all_ips_on_node
$node
454 for check
in $ips ; do
456 while read ip pnn
; do
457 if [ "$check" = "$ip" ] ; then
458 if [ "$pnn" = "$node" ] ; then
459 if $negating ; then return 1 ; fi
461 if ! $negating ; then return 1 ; fi
463 ips
="${ips/${ip}}" # Remove from list
466 # If we're negating and we didn't see the address then it
467 # isn't hosted by anyone!
469 ips
="${ips/${check}}"
471 done <<<"$out" # bashism to avoid problem setting variable in pipeline.
474 ips
="${ips// }" # Remove any spaces.
478 wait_until_ips_are_on_node
()
480 # Go to some trouble to print a use description of what is happening
482 if [ "$1" == "!" ] ; then
489 [ "$i" != "!" ] ||
continue
490 if [ -z "$node" ] ; then
494 ips
="${ips}${ips:+, }${i}"
496 echo "Waiting for ${ips} to ${not}be assigned to node ${node}"
498 wait_until
60 ips_are_on_node
"$@"
507 all_ips_on_node
$node
509 while read ip pnn
; do
510 if [ "$node" = "$pnn" ] ; then
513 done <<<"$out" # bashism to avoid problem setting variable in pipeline.
518 wait_until_node_has_some_ips
()
520 echo "Waiting for some IPs to be assigned to node ${test_node}"
522 wait_until
60 node_has_some_ips
"$@"
525 #######################################
527 _ctdb_hack_options
()
529 local ctdb_options
="$*"
531 case "$ctdb_options" in
532 *--start-as-stopped*)
533 export CTDB_START_AS_STOPPED
="yes"
539 _ctdb_hack_options
"$@"
541 if [ -e /etc
/redhat-release
] ; then
544 /etc
/init.d
/ctdb restart
548 # Restart CTDB on all nodes. Override for local daemons.
551 onnode
-p all
$CTDB_TEST_WRAPPER restart_ctdb_1
"$@"
554 # Nothing needed for a cluster. Override for local daemons.
562 # "$@" is passed to restart_ctdb_all.
564 echo -n "Restarting CTDB"
565 if $ctdb_test_restart_scheduled ; then
566 echo -n " (scheduled)"
571 for i
in $
(seq 1 5) ; do
572 _restart_ctdb_all
"$@" ||
{
573 echo "Restart failed. Trying again in a few seconds..."
578 wait_until_ready ||
{
579 echo "Cluster didn't become ready. Restarting..."
583 echo "Setting RerecoveryTimeout to 1"
584 onnode
-pq all
"$CTDB setvar RerecoveryTimeout 1"
586 # In recent versions of CTDB, forcing a recovery like this
587 # blocks until the recovery is complete. Hopefully this will
588 # help the cluster to stabilise before a subsequent test.
589 echo "Forcing a recovery..."
590 onnode
-q 0 $CTDB recover
593 if ! onnode
-q any
$CTDB_TEST_WRAPPER _cluster_is_recovered
; then
594 echo "Cluster has gone into recovery again, waiting..."
595 wait_until
30/2 onnode
-q any
$CTDB_TEST_WRAPPER _cluster_is_recovered
599 # Cluster is still healthy. Good, we're done!
600 if ! onnode
0 $CTDB_TEST_WRAPPER _cluster_is_healthy
; then
601 echo "Cluster became UNHEALTHY again [$(date)]"
602 onnode
-p all ctdb status
-X 2>&1
603 onnode
-p all ctdb scriptstatus
2>&1
608 echo "Doing a sync..."
609 onnode
-q 0 $CTDB sync
615 echo "Cluster UNHEALTHY... too many attempts..."
616 onnode
-p all ctdb status
-X 2>&1
617 onnode
-p all ctdb scriptstatus
2>&1
619 # Try to make the calling test fail
624 # Does nothing on cluster and should be overridden for local daemons
630 ctdb_restart_when_done
()
632 ctdb_test_restart_scheduled
=true
635 get_ctdbd_command_line_option
()
640 try_command_on_node
"$pnn" "$CTDB getpid" || \
641 die
"Unable to get PID of ctdbd on node $pnn"
643 local pid
="${out#*:}"
644 try_command_on_node
"$pnn" "ps -p $pid -o args hww" || \
645 die
"Unable to get command-line of PID $pid"
647 # Strip everything up to and including --option
648 local t
="${out#*--${option}}"
649 # Strip leading '=' or space if present
652 # Strip any following options and print
656 #######################################
658 wait_for_monitor_event
()
663 echo "Waiting for a monitor event on node ${pnn}..."
665 try_command_on_node
"$pnn" $CTDB scriptstatus ||
{
666 echo "Unable to get scriptstatus from node $pnn"
670 local ctdb_scriptstatus_original
="$out"
671 wait_until
120 _ctdb_scriptstatus_changed
674 _ctdb_scriptstatus_changed
()
676 try_command_on_node
"$pnn" $CTDB scriptstatus ||
{
677 echo "Unable to get scriptstatus from node $pnn"
681 [ "$out" != "$ctdb_scriptstatus_original" ]
684 #######################################
688 select_test_node_and_ips
690 nfs_first_export
=$
(showmount
-e $test_ip |
sed -n -e '2s/ .*//p')
692 echo "Creating test subdirectory..."
693 try_command_on_node
$test_node "mktemp -d --tmpdir=$nfs_first_export"
695 try_command_on_node
$test_node "chmod 777 $nfs_test_dir"
697 nfs_mnt_d
=$
(mktemp
-d)
698 nfs_local_file
="${nfs_mnt_d}/${nfs_test_dir##*/}/TEST_FILE"
699 nfs_remote_file
="${nfs_test_dir}/TEST_FILE"
701 ctdb_test_exit_hook_add nfs_test_cleanup
703 echo "Mounting ${test_ip}:${nfs_first_export} on ${nfs_mnt_d} ..."
704 mount
-o timeo
=1,hard
,intr
,vers
=3 \
705 "[${test_ip}]:${nfs_first_export}" ${nfs_mnt_d}
710 rm -f "$nfs_local_file"
711 umount -f "$nfs_mnt_d"
713 onnode -q $test_node rmdir "$nfs_test_dir"
716 #######################################
718 # If the given IP is hosted then print 2 items: maskbits and iface
724 *:*) _family="inet6
" ; _bits=128 ;;
725 *) _family="inet
" ; _bits=32 ;;
728 ip addr show to "${_addr}/${_bits}" 2>/dev/null | \
729 awk -v family="${_family}" \
730 'NR == 1 { iface = $2; sub(":$
", "", iface) } \
731 $1 ~ /inet/ { mask = $2; sub(".
*/", "", mask); \
732 print mask, iface, family }'
737 _addr="${1%/*}" # Remove optional maskbits
739 set -- $(ip_maskbits_iface $_addr)
740 if [ -n "$1" ] ; then
743 echo "Removing public address
$_addr/$_maskbits from device
$_iface"
744 ip addr del "$_ip/$_maskbits" dev "$_iface" >/dev/null 2>&1 || true
755 #######################################
757 # $1: pnn, $2: DB name
760 try_command_on_node -v $1 $CTDB getdbstatus "$2" |
761 sed -n -e "s@^path
: @@p
"
764 # $1: pnn, $2: DB name
765 db_ctdb_cattdb_count_records ()
767 try_command_on_node -v $1 $CTDB cattdb "$2" |
768 grep '^key' | grep -v '__db_sequence_number__' |
772 # $1: pnn, $2: DB name, $3: key string, $4: value string, $5: RSN (default 7)
775 _tdb=$(db_get_path $1 "$2")
777 try_command_on_node $1 $CTDB tstore "$_tdb" "$3" "$4" "$_rsn"
780 # $1: pnn, $2: DB name, $3: dbseqnum (must be < 255!!!!!)
781 db_ctdb_tstore_dbseqnum ()
783 # "__db_sequence_number__
" + trailing 0x00
784 _key='0x5f5f64625f73657175656e63655f6e756d6265725f5f00'
786 # Construct 8 byte (unit64_t) database sequence number. This
787 # probably breaks if $3 > 255
788 _value=$(printf "0x
%02x
%014x
" $3 0)
790 db_ctdb_tstore $1 "$2" "$_key" "$_value"
793 #######################################
795 # Make sure that $CTDB is set.
798 local="${TEST_SUBDIR}/scripts
/local.bash
"
799 if [ -r "$local" ] ; then