Context Navigation

Filesystem-qla-monitor @ 8397

Revision 8397, 23.7 KB (checked in by nlmills, 3 years ago)
initial merge with Orange-Branch. much will be broken

Line
1	#!/bin/sh
2	#
3	# Support: linux-ha@lists.linux-ha.org
4	# License: GNU General Public License (GPL)
5	#
6	# Filesystem
7	# Description: Manages a Filesystem on a shared storage medium.
8	# Original Author: Eric Z. Ayers (eric.ayers@compgen.com)
9	# Original Release: 25 Oct 2000
10	#
11	# Modified to support monitoring of a QLogic adapter, 2007
12	# Relies on underlying scripts named fs-instance-alarm.pl and
13	# fs-power-control.pl to take action in the event of monitoring failure
14	#
15	# usage: ./Filesystem {start\|stop\|status\|monitor\|validate-all\|meta-data}
16	#
17	# OCF parameters are as below:
18	# OCF_RESKEY_device
19	# OCF_RESKEY_directory
20	# OCF_RESKEY_fstype
21	# OCF_RESKEY_options
22	# OCF_RESKEY_fsname
23	# OCF_RESKEY_conf_dir
24	#
25	#OCF_RESKEY_device : name of block device for the filesystem. e.g. /dev/sda1, /dev/md0
26	# Or a -U or -L option for mount, or an NFS mount specification
27	#OCF_RESKEY_directory : the mount point for the filesystem
28	#OCF_RESKEY_fstype : optional name of the filesystem type. e.g. ext2
29	#OCF_RESKEY_options : options to be given to the mount command via -o
30	#OCF_RESKEY_fsname : file system name (PVFS2)
31	#OCF_RESKEY_conf_dir : file system conf directory (PVFS2)
32	#
33	#
34	# An example usage in /etc/ha.d/haresources:
35	# node1 10.0.0.170 Filesystem::/dev/sda1::/data1::ext2
36	# or
37	# node1 10.0.0.170 Filesystem::-Ldata1::/data1::ext2
38	# or
39	# node1 10.0.0.170 Filesystem::server:/data1::/data1::nfs::ro
40	#
41	# This assumes you want to manage a filesystem on a shared (scsi) bus.
42	# Do not put this filesystem in /etc/fstab. This script manages all of
43	# that for you.
44	#
45	# If you are interested in High Availability, you will probably also want
46	# some sort of external hardware RAID controller in front of the actual
47	# disks. I don't mean a RAID controller embedded in the host controller -
48	# it has to be an external controller.
49	#
50	# It can also be an internal RAID controller if the controller supports
51	# failover. IBM's ServeRAID controller does this, and it automatically
52	# prohibits concurrent access too, so it's pretty cool in this application.
53	#
54	# There is a script for software RAID-1 included in this directory. Right
55	# now, I wouldn't recommend using software RAID (see notes in the Raid1 script)
56	#
57	# NOTE: There is no locking (such as a SCSI reservation) being done here.
58	# I would if the SCSI driver could properly maintain the reservation,
59	# which it cannot, even with the 'scsi reservation' patch submitted
60	# earlier this year by James Bottomley. The patch minimizes the
61	# bus resets caused by a RESERVATION_CONFLICT return, and helps the
62	# reservation stay when 2 nodes contend for a reservation,
63	# but it does not attempt to recover the reservation in the
64	# case of a bus reset.
65	#
66	# What all this means is that if 2 nodes mount the same file system
67	# read-write, the filesystem is going to become corrupted.
68	#
69	# As a result, you should use this together with the stonith option
70	# and redundant, independent communications paths.
71	#
72	# If you don't do this, don't blame us when you scramble your disk.
73	#
74	# Note: the ServeRAID controller does prohibit concurrent acess
75	# In this case, you don't actually need STONITH, but redundant comm is
76	# still an excellent idea.
77	#
78
79	#######################################################################
80	# Initialization:
81
82	# newer versions of heartbeat have moved the ocf-shellfuncs file
83	if [ -f /usr/lib/ocf/resource.d/heartbeat/.ocf-shellfuncs ] ; then
84	. /usr/lib/ocf/resource.d/heartbeat/.ocf-shellfuncs
85	else
86	. /usr/lib/heartbeat/ocf-shellfuncs
87	fi
88
89	#######################################################################
90
91	# Utilities used by this script
92	MODPROBE=/sbin/modprobe
93	FSCK=/sbin/fsck
94	FUSER=/sbin/fuser
95	MOUNT=/bin/mount
96	UMOUNT=/bin/umount
97	BLOCKDEV=/sbin/blockdev
98
99	check_util () {
100	if [ ! -x "$1" ] ; then
101	ocf_log err "Setup problem: Couldn't find utility $1"
102	exit $OCF_ERR_GENERIC
103	fi
104	}
105
106	usage() {
107	cat <<-EOT
108	usage: $0 {start\|stop\|status\|monitor\|validate-all\|meta-data}
109	EOT
110	}
111
112	meta_data() {
113	cat <<END
114	<?xml version="1.0"?>
115	<!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
116	<resource-agent name="Filesystem">
117	<version>1.0</version>
118
119	<longdesc lang="en">
120	Resource script for Filesystem. It manages a Filesystem on a shared storage medium.
121	</longdesc>
122	<shortdesc lang="en">Filesystem resource agent</shortdesc>
123
124	<parameters>
125	<parameter name="device" unique="0" required="1">
126	<longdesc lang="en">
127	The name of block device for the filesystem, or -U, -L options for mount, or NFS mount specification.
128	</longdesc>
129	<shortdesc lang="en">block device</shortdesc>
130	<content type="string" default="" />
131	</parameter>
132
133	<parameter name="directory" unique="0" required="1">
134	<longdesc lang="en">
135	The mount point for the filesystem.
136	</longdesc>
137	<shortdesc lang="en">mount point</shortdesc>
138	<content type="string" default="" />
139	</parameter>
140
141	<parameter name="fsname" unique="0" required="1">
142	<longdesc lang="en">
143	Name of PVFS2 file system
144	</longdesc>
145	<shortdesc lang="en">File system</shortdesc>
146	<content type="string" default="" />
147	</parameter>
148
149	<parameter name="conf_dir" unique="0" required="1">
150	<longdesc lang="en">
151	Path to file system conf directory
152	</longdesc>
153	<shortdesc lang="en">File system conf directory</shortdesc>
154	<content type="string" default="" />
155	</parameter>
156
157	<parameter name="fstype" unique="0" required="1">
158	<longdesc lang="en">
159	The optional type of filesystem to be mounted.
160	</longdesc>
161	<shortdesc lang="en">filesystem type</shortdesc>
162	<content type="string" default="" />
163	</parameter>
164
165	<parameter name="options" unique="0">
166	<longdesc lang="en">
167	Any extra options to be given as -o options to mount.
168	</longdesc>
169	<shortdesc lang="en">options</shortdesc>
170	<content type="string" default="" />
171	</parameter>
172
173	<parameter name="ocfs2_cluster" unique="0">
174	<longdesc lang="en">
175	The name (UUID) of the OCFS2 cluster this filesystem is part of,
176	iff this is an OCFS2 resource and there's more than one cluster. You
177	should not need to specify this.
178	</longdesc>
179	<shortdesc lang="en">OCFS2 cluster name/UUID</shortdesc>
180	<content type="string" default="" />
181	</parameter>
182
183	<parameter name="ocfs2_configfs" unique="0">
184	<longdesc lang="en">
185	Mountpoint of the cluster hierarchy below configfs. You should not
186	need to specify this.
187	</longdesc>
188	<shortdesc lang="en">OCFS2 configfs root</shortdesc>
189	<content type="string" default="" />
190	</parameter>
191	</parameters>
192
193	<actions>
194	<action name="start" timeout="60" />
195	<action name="stop" timeout="60" />
196	<action name="notify" timeout="60" />
197	<action name="status" depth="0" timeout="10" interval="10" start-delay="10" />
198	<action name="monitor" depth="0" timeout="10" interval="10" start-delay="10" />
199	<action name="validate-all" timeout="5" />
200	<action name="meta-data" timeout="5" />
201	</actions>
202	</resource-agent>
203	END
204	}
205
206	#
207	# Make sure the kernel does the right thing with the FS buffers
208	# This function should be called after unmounting and before mounting
209	# It may not be necessary in 2.4 and later kernels, but it shouldn't hurt
210	# anything either...
211	#
212	# It's really a bug that you have to do this at all...
213	#
214	flushbufs() {
215	if
216	[ "$BLOCKDEV" != "" -a -x "$BLOCKDEV" -a "$blockdevice" = "yes" ]
217	then
218	$BLOCKDEV --flushbufs $1
219	return $?
220	fi
221
222	return 0
223	}
224
225	# Take advantage of /proc/mounts if present, use portabel mount command
226	# otherwise. Normalize format to "dev mountpoint fstype".
227	list_mounts() {
228	if [ -f "/proc/mounts" -a -r "/proc/mounts" ]; then
229	cut -d' ' -f1,2,3 </proc/mounts
230	else
231	$MOUNT \| cut -d' ' -f1,3,5
232	fi
233	}
234
235	determine_blockdevice() {
236	if [ $blockdevice == "yes" ]; then
237	return
238	fi
239
240	# Get the current real device name, if possible.
241	# (specified devname could be -L or -U...)
242	case "$FSTYPE" in
243	nfs\|smbfs\|cifs) ;;
244	*) DEVICE=`list_mounts \| grep " $MOUNTPOINT " \| cut -d' ' -f1`
245	if [ -b "$DEVICE" ]; then
246	blockdevice=yes
247	fi
248	;;
249	esac
250	}
251
252	# Lists all filesystems potentially mounted under a given path,
253	# excluding the path itself.
254	list_submounts() {
255	list_mounts \| grep " $1/" \| cut -d' ' -f2 \| sort -r
256	}
257
258	ocfs2_cleanup() {
259	# We'll never see the post-stop notification. We're gone now,
260	# have unmounted, and thus should remove the membership.
261	#
262	# (Do so regardless of whether we were unmounted already,
263	# because the admin might have manually unmounted but not
264	# cleared up the membership directory. Bad admin, no cookie.)
265	#
266	if [ ! -d "$OCFS2_FS_ROOT" ]; then
267	ocf_log info "$OCFS2_FS_ROOT: Filesystem membership already gone."
268	else
269	ocf_log info "$OCFS2_FS_ROOT: Removing membership directory."
270	rm -rf $OCFS2_FS_ROOT/
271	fi
272	}
273
274	ocfs2_init()
275	{
276	# Check & initialize the OCFS2 specific variables.
277	if [ -z "$OCF_RESKEY_CRM_meta_clone_max" ]; then
278	ocf_log err "ocfs2 must be run as a clone."
279	exit $OCF_ERR_GENERIC
280	fi
281
282	if [ $blockdevice = "no" ]; then
283	ocf_log err "$DEVICE: ocfs2 needs a block device instead."
284	exit $OCF_ERR_GENERIC
285	fi
286
287	for f in "$OCF_RESKEY_ocfs2_configfs" /sys/kernel/config/cluster /configfs/cluster ; do
288	if [ -n "$f" -a -d "$f" ]; then
289	OCFS2_CONFIGFS="$f"
290	break
291	fi
292	done
293	if [ ! -d "$OCFS2_CONFIGFS" ]; then
294	ocf_log err "ocfs2 needs configfs mounted."
295	exit $OCF_ERR_GENERIC
296	fi
297
298	OCFS2_UUID=$(mounted.ocfs2 -d $DEVICE\|tail -1\|awk '{print $3}'\|tr -d -- -\|tr a-z A-Z)
299	if [ -z "$OCFS2_UUID" -o "$OCFS2_UUID" = "UUID" ]; then
300	ocf_log err "$DEVICE: Could not determine ocfs2 UUID."
301	exit $OCF_ERR_GENERIC
302	fi
303
304	if [ -n "$OCF_RESKEY_ocfs2_cluster" ]; then
305	OCFS2_CLUSTER=$(echo $OCF_RESKEY_ocfs2_cluster \| tr a-z A-Z)
306	else
307	OCFS2_CLUSTER=$(find "$OCFS2_CONFIGFS" -maxdepth 1 -mindepth 1 -type d -printf %f 2>/dev/null)
308	set -- $OCFS2_CLUSTER
309	local n="$#"
310	if [ $n -gt 1 ]; then
311	ocf_log err "$OCFS2_CLUSTER: several clusters found."
312	exit $OCF_ERR_GENERIC
313	fi
314	if [ $n -eq 0 ]; then
315	ocf_log err "$OCFS2_CONFIGFS: no clusters found."
316	exit $OCF_ERR_GENERIC
317	fi
318	fi
319
320	OCFS2_CLUSTER_ROOT="$OCFS2_CONFIGFS/$OCFS2_CLUSTER"
321	if [ ! -d "$OCFS2_CLUSTER_ROOT" ]; then
322	ocf_log err "$OCFS2_CLUSTER: Cluster doesn't exist. Maybe o2cb hasn't been run?"
323	exit $OCF_ERR_GENERIC
324	fi
325
326	OCFS2_FS_ROOT=$OCFS2_CLUSTER_ROOT/heartbeat/$OCFS2_UUID
327	}
328
329	#
330	# START: Start up the filesystem
331	#
332	Filesystem_start()
333	{
334	if [ "$FSTYPE" = "ocfs2" ]; then
335	# "start" now has the notification data available; that
336	# we're being started means we didn't get the
337	# pre-notification, because we weren't running, so
338	# process the information now first.
339	ocf_log info "$OCFS2_UUID: Faking pre-notification on start."
340	OCF_RESKEY_CRM_meta_notify_type="pre"
341	OCF_RESKEY_CRM_meta_notify_operation="start"
342	Filesystem_notify
343	fi
344
345	# See if the device is already mounted.
346	if Filesystem_status >/dev/null 2>&1 ; then
347	ocf_log info "Filesystem $MOUNTPOINT is already mounted."
348	return $OCF_SUCCESS
349	fi
350
351	# Insert SCSI module
352	# TODO: This probably should go away. Why should the filesystem
353	# RA magically load a kernel module?
354	$MODPROBE scsi_hostadapter >/dev/null 2>&1
355
356	if [ -z $FSTYPE ]; then
357	: No $FSTYPE specified, rely on the system has the right file-system support already
358	else
359	# Insert Filesystem module
360	$MODPROBE $FSTYPE >/dev/null 2>&1
361	grep -e "$FSTYPE"'$' /proc/filesystems >/dev/null
362	if [ $? != 0 ] ; then
363	ocf_log err "Couldn't find filesystem $FSTYPE in /proc/filesystems"
364	return $OCF_ERR_ARGS
365	fi
366	fi
367
368	# Check the filesystem & auto repair.
369	# NOTE: Some filesystem types don't need this step... Please modify
370	# accordingly
371
372	if [ $blockdevice = "yes" ]; then
373	if [ ! -b "$DEVICE" ] ; then
374	ocf_log err "Couldn't find device [$DEVICE]. Expected /dev/??? to exist"
375	exit $OCF_ERR_ARGS
376	fi
377
378	if
379	case $FSTYPE in
380	ext3\|reiserfs\|reiser4\|nss\|xfs\|jfs\|vfat\|fat\|nfs\|cifs\|smbfs\|ocfs2) false;;
381	*) true;;
382	esac
383	then
384	ocf_log info "Starting filesystem check on $DEVICE"
385	if [ -z $FSTYPE ]; then
386	$FSCK -a $DEVICE
387	else
388	$FSCK -t $FSTYPE -a $DEVICE
389	fi
390
391	# NOTE: if any errors at all are detected, it returns non-zero
392	# if the error is >= 4 then there is a big problem
393	if
394	[ $? -ge 4 ]
395	then
396	ocf_log err "Couldn't sucessfully fsck filesystem for $DEVICE"
397	return $OCF_ERR_GENERIC
398	fi
399	fi
400	fi
401
402	if [ ! -d "$MOUNTPOINT" ] ; then
403	ocf_log err "Couldn't find directory [$MOUNTPOINT] to use as a mount point"
404	exit $OCF_ERR_ARGS
405	fi
406
407	flushbufs $DEVICE
408	# Mount the filesystem.
409	if [ -z $FSTYPE ]; then
410	$MOUNT $options $DEVICE $MOUNTPOINT
411	else
412	$MOUNT -t $FSTYPE $options $DEVICE $MOUNTPOINT
413	fi
414
415	if [ $? -ne 0 ]; then
416	ocf_log err "Couldn't mount filesystem $DEVICE on $MOUNTPOINT"
417	if [ "$FSTYPE" = "ocfs2" ]; then
418	ocfs2_cleanup
419	fi
420	return $OCF_ERR_GENERIC
421	fi
422
423	return 0
424	}
425	# end of Filesystem_start
426
427	Filesystem_notify() {
428	# Process notifications; this is the essential glue level for
429	# giving user-space membership events to a cluster-aware
430	# filesystem. Right now, only OCFS2 is supported.
431	#
432	# When we get a pre-start notification, we set up all the nodes
433	# which will be active in our membership for the filesystem.
434	# (For the resource to be started, this happens at the time of
435	# the actual 'start' operation.)
436	#
437	# At a post-start, actually there's nothing to do for us really,
438	# but no harm done in re-syncing either.
439	#
440	# pre-stop is meaningless; we can't remove any node yet, it
441	# first needs to unmount.
442	#
443	# post-stop: the node is removed from the membership of the
444	# other nodes.
445	#
446	# Note that this expects that the base cluster is already
447	# active; ie o2cb has been started and populated
448	# $OCFS2_CLUSTER_ROOT/node/ already. This can be achieved by
449	# simply having o2cb run on all nodes by the CRM too. This
450	# probably ought to be mentioned somewhere in the to be written
451	# documentation. ;-)
452	#
453
454	if [ "$FSTYPE" != "ocfs2" ]; then
455	# One of the cases which shouldn't occur; it should have
456	# been caught much earlier. Still, you know ...
457	ocf_log err "$DEVICE: Notification received for non-ocfs2 mount."
458	return $OCF_ERR_GENERIC
459	fi
460
461	local n_type="$OCF_RESKEY_CRM_meta_notify_type"
462	local n_op="$OCF_RESKEY_CRM_meta_notify_operation"
463	local n_active="$OCF_RESKEY_CRM_meta_notify_active_uname"
464	local n_stop="$OCF_RESKEY_CRM_meta_notify_stop_uname"
465	local n_start="$OCF_RESKEY_CRM_meta_notify_start_uname"
466
467	ocf_log info "$OCFS2_UUID: notify: $n_type for $n_op"
468	ocf_log info "$OCFS2_UUID: notify active: $n_active"
469	ocf_log info "$OCFS2_UUID: notify stop: $n_stop"
470	ocf_log info "$OCFS2_UUID: notify start: $n_start"
471
472	case "$n_type" in
473	pre)
474	case "$n_op" in
475	stop)
476	ocf_log info "$OCFS2_UUID: ignoring pre-notify for stop."
477	return $infoUCCESS
478	;;
479	start)
480	# These are about to become active; prepare to
481	# communicate with them.
482	n_active="$n_active $n_start"
483	;;
484	esac
485	;;
486	post)
487	case "$n_op" in
488	stop)
489	# remove unames from notify_stop_uname; these have been
490	# stopped and can no longer be considered active.
491	for UNAME in "$n_stop"; do
492	n_active="${n_active//$UNAME/}"
493	done
494	;;
495	start)
496	if [ "$n_op" = "start" ]; then
497	ocf_log info "$OCFS2_UUID: ignoring post-notify for start."
498	return $OCF_SUCCESS
499	fi
500	;;
501	esac
502	;;
503	esac
504
505	ocf_log info "$OCFS2_UUID: post-processed active: $n_active"
506
507	local n_myself=${HA_CURHOST:-$(uname -n \| tr A-Z a-z)}
508	ocf_log info "$OCFS2_UUID: I am node $n_myself."
509
510	case " $n_active " in
511	" $n_myself ") ;;
512	*) ocf_log err "$OCFS2_UUID: $n_myself (local) not on active list!"
513	return $OCF_ERR_GENERIC
514	;;
515	esac
516
517	if [ -d "$OCFS2_FS_ROOT" ]; then
518	entry_prefix=$OCFS2_FS_ROOT/
519	for entry in $OCFS2_FS_ROOT/* ; do
520	n_fs="${entry##$entry_prefix}"
521	# ocf_log info "$OCFS2_UUID: Found current node $n_fs"
522	case " $n_active " in
523	" $n_fs ")
524	# Construct a list of nodes which are present
525	# already in the membership.
526	n_exists="$n_exists $n_fs"
527	ocf_log info "$OCFS2_UUID: Keeping node: $n_fs"
528	;;
529	*)
530	# Node is in the membership currently, but not on our
531	# active list. Must be removed.
532	if [ "$n_op" = "start" ]; then
533	ocf_log warn "$OCFS2_UUID: Removing nodes on start"
534	fi
535	ocf_log info "$OCFS2_UUID: Removing dead node: $n_fs"
536	if ! rm -f $entry ; then
537	ocf_log err "$OCFS2_UUID: Removal of $n_fs failed!"
538	fi
539	;;
540	esac
541	done
542	else
543	ocf_log info "$OCFS2_UUID: heartbeat directory doesn't exist yet, creating."
544	mkdir -p $OCFS2_FS_ROOT
545	fi
546
547	ocf_log info "$OCFS2_UUID: Existing node list: $n_exists"
548
549	# (2)
550	for entry in $n_active ; do
551	# ocf_log info "$OCFS2_UUID: Expected active node: $entry"
552	case " $n_exists " in
553	" $entry ")
554	ocf_log info "$OCFS2_UUID: Already active: $entry"
555	;;
556	*)
557	if [ "$n_op" = "stop" ]; then
558	ocf_log warn "$OCFS2_UUID: Adding nodes on stop"
559	fi
560	ocf_log info "$OCFS2_UUID: Activating node: $entry"
561	if ! ln -s $OCFS2_CLUSTER_ROOT/node/$entry $OCFS2_FS_ROOT/$entry ; then
562	ocf_log err "$OCFS2_CLUSTER_ROOT/node/$entry: failed to link"
563	fi
564	;;
565	esac
566	done
567	}
568
569	#
570	# STOP: Unmount the filesystem
571	#
572	Filesystem_stop()
573	{
574	# See if the device is currently mounted
575	Filesystem_status >/dev/null 2>&1
576	if [ $? -ne $OCF_NOT_RUNNING ]; then
577	# Determine the real blockdevice this is mounted on (if
578	# possible) prior to unmounting.
579	determine_blockdevice
580
581	# For networked filesystems, there's merit in trying -f:
582	case "$FSTYPE" in
583	nfs\|cifs\|smbfs) umount_force="-f" ;;
584	esac
585
586	# Umount all sub-filesystems mounted under $MOUNTPOINT/ too.
587	for SUB in `list_submounts $MOUNTPOINT` $MOUNTPOINT; do
588	ocf_log info "Trying to unmount $MOUNTPOINT"
589	for sig in SIGTERM SIGTERM SIGTERM SIGKILL SIGKILL SIGKILL; do
590	if $UMOUNT $umount_force $SUB ; then
591	rc=$OCF_SUCCESS
592	ocf_log info "unmounted $SUB successfully"
593	break
594	else
595	rc=$OCF_ERR_GENERIC
596	ocf_log err "Couldn't unmount $SUB; trying cleanup with $sig"
597	# fuser returns a non-zero return code if none of the
598	# specified files is accessed or in case of a fatal
599	# error.
600	if $FUSER -$sig -m -k $SUB ; then
601	ocf_log info "Some processes on $SUB were signalled"
602	else
603	ocf_log info "No processes on $SUB were signalled"
604	fi
605	sleep 1
606	fi
607	done
608
609	if [ $rc -ne $OCF_SUCCESS ]; then
610	ocf_log err "Couldn't unmount $SUB, giving up!"
611	fi
612	done
613	else
614	# Already unmounted, wonderful.
615	rc=$OCF_SUCCESS
616	fi
617
618	flushbufs $DEVICE
619
620	if [ "$FSTYPE" = "ocfs2" ]; then
621	ocfs2_cleanup
622	fi
623
624	return $rc
625	}
626	# end of Filesystem_stop
627
628	#
629	# STATUS: is the file system mounted and the SAN path alive?
630	#
631	Filesystem_status_qla_monitor()
632	{
633	Filesystem_status >/dev/null 2>&1
634	ORIG_RET=$?
635	if [ $ORIG_RET -eq $OCF_SUCCESS ]; then
636	# the fs is mounted. Are the paths ok?
637	/usr/bin/qla-monitor.pl --device $DEVICE
638	if [ $? -ne 0 ]; then
639	# log as much as we can
640	ocf_log info "Error: $DEVICE is dead."
641	echo "Error: $DEVICE is dead on node `hostname`, failing over" >> /var/log/pvfs2/pvfs2-failover.log;
642	/usr/bin/fs-instance-alarm.pl --fs-name $fsname --ce `hostname` --type PVFS2_HA --msg "$DEVICE is dead, failing over."
643	/usr/bin/fs-power-control.pl --fs-name $fsname --conf-dir $conf_dir --host `hostname` --command reboot
644	# shoot ourselves in the head
645	return $OCF_ERR_GENERIC
646	fi
647	dd bs=4096 count=1 if=/dev/zero oflag=direct of=${MOUNTPOINT}/filesystem.monitor
648	if [ $? -ne 0 ]; then
649	ocf_log err "Error: Unable to write to $DEVICE on node `hostname`."
650	/usr/bin/logger -p user.err -t PVFS2 "Error: Unable to write to $DEVICE on node `hostname`, failing over."
651	/usr/bin/fs-instance-alarm.pl --fs-name $fsname --ce `hostname` --type PVFS2_HA --msg "Error: Unable to write to $DEVICE on node `hostname`, failing over."
652	# shoot ourselves in the head
653	/usr/bin/fs-power-control.pl --fs-name $fsname --conf-dir $conf_dir --host `hostname` --command reboot
654	return $OCF_ERR_GENERIC
655	fi
656	fi
657	return $ORIG_RET
658	}
659	# end of Filesystem_status_qla_monitor
660
661	#
662	# STATUS: is the filesystem mounted or not?
663	#
664	Filesystem_status()
665	{
666	if
667	list_mounts \| grep -q " $MOUNTPOINT " >/dev/null 2>&1
668	then
669	rc=$OCF_SUCCESS
670	msg="$MOUNTPOINT is mounted (running)"
671	else
672	rc=$OCF_NOT_RUNNING
673	msg="$MOUNTPOINT is unmounted (stopped)"
674	fi
675
676	# TODO: For ocfs2, or other cluster filesystems, should we be
677	# checking connectivity to other nodes here, or the IO path to
678	# the storage?
679
680	case "$OP" in
681	status) ocf_log info "$msg";;
682	esac
683	return $rc
684	}
685	# end of Filesystem_status
686
687	#
688	# VALIDATE_ALL: Are the instance parameters valid?
689	# FIXME!! The only part that's useful is the return code.
690	# This code always returns $OCF_SUCCESS (!)
691	#
692	Filesystem_validate_all()
693	{
694	if [ -n $MOUNTPOINT -a ! -d $MOUNTPOINT ]; then
695	ocf_log warn "Mountpoint $MOUNTPOINT does not exist"
696	fi
697
698	# Check if the $FSTYPE is workable
699	# NOTE: Without inserting the $FSTYPE module, this step may be imprecise
700	# TODO: This is Linux specific crap.
701	if [ ! -z $FSTYPE ]; then
702	cut -f2 /proc/filesystems \|grep -q ^$FSTYPE$
703	if [ $? -ne 0 ]; then
704	modpath=/lib/modules/`uname -r`
705	moddep=$modpath/modules.dep
706	# Do we have $FSTYPE in modules.dep?
707	cut -d' ' -f1 $moddep \|grep -q "^$modpath.*$FSTYPE\.k\?o:$"
708	if [ $? -ne 0 ]; then
709	ocf_log info "It seems we do not have $FSTYPE support"
710	fi
711	fi
712	fi
713
714	#TODO: How to check the $options ?
715
716	return $OCF_SUCCESS
717	}
718
719	# Check the arguments passed to this script
720	if
721	[ $# -ne 1 ]
722	then
723	usage
724	exit $OCF_ERR_ARGS
725	fi
726
727	OP=$1
728
729	# These operations do not require instance parameters
730	case $OP in
731	meta-data) meta_data
732	exit $OCF_SUCCESS
733	;;
734	usage) usage
735	exit $OCF_SUCCESS
736	;;
737	esac
738
739	if
740	[ -z "$OCF_RESKEY_fsname" ]
741	then
742	ocf_log err "Filesystem-qla-monitor must specify fsname!"
743	# usage
744	exit $OCF_ERR_GENERIC
745	fi
746
747	if
748	[ -z "$OCF_RESKEY_conf_dir" ]
749	then
750	ocf_log err "Filesystem-qla-monitor must specify conf_dir!"
751	# usage
752	exit $OCF_ERR_GENERIC
753	fi
754
755	# Check the OCF_RESKEY_ environment variables...
756	DEVICE=$OCF_RESKEY_device
757	FSTYPE=$OCF_RESKEY_fstype
758	if [ ! -z "$OCF_RESKEY_options" ]; then
759	options="-o $OCF_RESKEY_options"
760	fi
761	fsname=$OCF_RESKEY_fsname
762	conf_dir=$OCF_RESKEY_conf_dir
763
764	blockdevice=no
765	case $DEVICE in
766	"") ocf_log err "Please set OCF_RESKEY_device to the device to be managed"
767	exit $OCF_ERR_ARGS
768	;;
769	-*) # Oh... An option to mount instead... Typically -U or -L
770	;;
771	[^/]:/) # An NFS filesystem specification...
772	;;
773	//[^/]/) # An SMB filesystem specification...
774	;;
775	*) if [ ! -b "$DEVICE" -a "X$OP" != Xstart ] ; then
776	ocf_log warn "Couldn't find device [$DEVICE]. Expected /dev/??? to exist"
777	fi
778	blockdevice=yes
779	;;
780	esac
781
782	case $FSTYPE in
783	ocfs2)
784	ocfs2_init
785	;;
786	nfs)
787	: # this is kind of safe too
788	;;
789	*)
790	if [ -n "$OCF_RESKEY_CRM_meta_clone_max" ]; then
791	ocf_log err "DANGER! $FSTYPE on $DEVICE is NOT cluster-aware!"
792	ocf_log err "DO NOT RUN IT AS A CLONE!"
793	ocf_log err "Politely refusing to proceed to avoid data corruption."
794	exit $OCF_ERR_GENERIC
795	fi
796	;;
797	esac
798
799	# It is possible that OCF_RESKEY_directory has one or even multiple trailing "/".
800	# But the output of `mount` and /proc/mounts do not.
801	if [ -z $OCF_RESKEY_directory ]; then
802	if [ X$OP = "Xstart" -o $blockdevice = "no" ]; then
803	ocf_log err "Please specify the directory"
804	exit $OCF_ERR_ARGS
805	fi
806	else
807	MOUNTPOINT=$(echo $OCF_RESKEY_directory \| sed 's/\/*$//')
808	: ${MOUNTPOINT:=/}
809	# At this stage, $MOUNTPOINT does not contain trailing "/" unless it is "/"
810	# TODO: / mounted via Filesystem sounds dangerous. On stop, we'll
811	# kill the whole system. Is that a good idea?
812	fi
813
814	# Check to make sure the utilites are found
815	check_util $MODPROBE
816	check_util $FSCK
817	check_util $FUSER
818	check_util $MOUNT
819	check_util $UMOUNT
820
821	if [ "$OP" != "monitor" ]; then
822	ocf_log info "Running $OP for $DEVICE on $MOUNTPOINT"
823	fi
824
825	case $OP in
826	start) Filesystem_start
827	;;
828	notify) Filesystem_notify
829	;;
830	stop) Filesystem_stop
831	;;
832	status\|monitor) Filesystem_status_qla_monitor
833	;;
834	validate-all) Filesystem_validate_all
835	;;
836	*) usage
837	exit $OCF_ERR_UNIMPLEMENTED
838	;;
839	esac
840	exit $?

Note: See TracBrowser for help on using the browser.

Context Navigation

root/branches/cu-security-branch/examples/heartbeat/hardware-specific/Filesystem-qla-monitor @ 8397

Download in other formats: