Просмотр исходного кода

Add monitoring and watchdog services.

git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@3053 fd59a12c-fef9-0310-b244-a6a79926bd2f
Angus Salkeld 15 лет назад
Родитель
Сommit
07d06c0c0f
8 измененных файлов с 1462 добавлено и 2 удалено
  1. 36 0
      configure.ac
  2. 14 0
      corosync.spec.in
  3. 1 1
      exec/Makefile.am
  4. 12 0
      exec/service.c
  5. 3 1
      include/corosync/corodefs.h
  6. 6 0
      services/Makefile.am
  7. 635 0
      services/mon.c
  8. 755 0
      services/wd.c

+ 36 - 0
configure.ac

@@ -252,6 +252,16 @@ AC_ARG_ENABLE([rdma],
 	[ enable_rdma="no" ])
 AM_CONDITIONAL(BUILD_RDMA, test x$enable_rdma = xyes)
 
+AC_ARG_ENABLE([monitoring],
+	[  --enable-monitoring              : resource monitoring ],,
+	[ default="no" ])
+AM_CONDITIONAL(BUILD_MONITORING, test x$enable_monitoring = xyes)
+
+AC_ARG_ENABLE([watchdog],
+	[  --enable-watchdog                   : Watchdog support ],,
+	[ edefault="no" ])
+AM_CONDITIONAL(BUILD_WATCHDOG, test x$enable_watchdog = xyes)
+
 AC_ARG_ENABLE([augeas],
 	[  --enable-augeas                 : Install the augeas lens for corosync.conf ],,
 	[ enable_augeas="no" ])
@@ -277,6 +287,7 @@ AC_ARG_WITH([socket-dir],
 # THIS SECTION MUST DIE!
 CP=cp
 OS_LDL="-ldl"
+have_linux="no"
 case "$host_os" in
 	*linux*)
 		AC_DEFINE_UNQUOTED([COROSYNC_LINUX], [1],
@@ -286,6 +297,7 @@ case "$host_os" in
 		OS_LDFLAGS=""
 		OS_DYFLAGS="-rdynamic"
 		DARWIN_OPTS=""
+		have_linux="yes"
 	;;
 	darwin*)
 		AC_DEFINE_UNQUOTED([COROSYNC_DARWIN], [1],
@@ -387,6 +399,30 @@ if test "x${enable_rdma}" = xyes; then
 	PACKAGE_FEATURES="$PACKAGE_FEATURES rdma"
 fi
 
+if test "x${enable_monitoring}" = xyes; then
+
+	AC_CHECK_LIB([statgrab], [sg_get_mem_stats], have_libstatgrab="yes", have_libstatgrab="no")
+
+	if test "x${have_libstatgrab}" = xyes; then
+		AC_DEFINE_UNQUOTED([HAVE_LIBSTATGRAB], 1, [have libstatgrab])
+		statgrab_LIBS="-lstatgrab"
+	else
+		if test "x${have_linux}" = xno; then
+			AC_MSG_ERROR(monitoring requires libstatgrab on non-linux systems)
+		fi
+	fi
+	AC_SUBST([statgrab_LIBS])
+	AC_DEFINE_UNQUOTED([HAVE_MONITORING], 1, [have resource monitoring])
+	PACKAGE_FEATURES="$PACKAGE_FEATURES monitoring"
+fi
+
+if test "x${enable_watchdog}" = xyes; then
+	AC_CHECK_HEADER(linux/watchdog.h,,AC_MSG_ERROR(watchdog requires linux/watchdog.h))
+	AC_CHECK_HEADER(linux/reboot.h,,AC_MSG_ERROR(watchdog requires linux/reboot.h))
+	AC_DEFINE_UNQUOTED([HAVE_WATCHDOG], 1, [have watchdog])
+	PACKAGE_FEATURES="$PACKAGE_FEATURES watchdog"
+fi
+
 if test "x${enable_augeas}" = xyes; then
 	PACKAGE_FEATURES="$PACKAGE_FEATURES augeas"
 fi

+ 14 - 0
corosync.spec.in

@@ -5,6 +5,8 @@
 # Invoke "rpmbuild --without <feature>" or "rpmbuild --with <feature>"
 # to disable or enable specific features
 %bcond_with testagents
+%bcond_with watchdog
+%bcond_with monitoring
 
 Name: corosync
 Summary: The Corosync Cluster Engine and Application Programming Interfaces
@@ -51,6 +53,12 @@ export rdmacm_LIBS=-lrdmacm \
 	--enable-nss \
 %if %{with testagents}
 	--enable-testagents \
+%endif
+%if %{with watchdog}
+	--enable-watchdog \
+%endif
+%if %{with monitoring}
+	--enable-monitoring \
 %endif
 	--enable-rdma \
 	--with-initddir=%{_initrddir}
@@ -115,6 +123,12 @@ fi
 %{_libexecdir}/lcrso/quorum_testquorum.lcrso
 %{_libexecdir}/lcrso/vsf_quorum.lcrso
 %{_libexecdir}/lcrso/vsf_ykd.lcrso
+%if %{with watchdog}
+%{_libexecdir}/lcrso/service_wd.lcrso
+%endif
+%if %{with monitoring}
+%{_libexecdir}/lcrso/service_mon.lcrso
+%endif
 %dir %{_localstatedir}/lib/corosync
 %dir %{_localstatedir}/log/cluster
 %{_mandir}/man8/corosync_overview.8*

+ 1 - 1
exec/Makefile.am

@@ -59,7 +59,7 @@ libcoroipcs_a_SOURCES	= $(COROIPCS_SRC)
 corosync_SOURCES 	= main.c util.c sync.c apidef.c service.c \
 			  timer.c totemconfig.c mainconfig.c quorum.c schedwrk.c \
 			  ../lcr/lcr_ifact.c evil.c syncv2.c
-corosync_LDADD	  	= -ltotem_pg -llogsys -lcoroipcs
+corosync_LDADD	  	= -ltotem_pg -llogsys -lcoroipcs $(statgrab_LIBS)
 corosync_DEPENDENCIES	= libtotem_pg.so.$(SONAME) liblogsys.so.$(SONAME) libcoroipcs.so.$(SONAME)
 corosync_LDFLAGS	= $(OS_DYFLAGS) -L./
 

+ 12 - 0
exec/service.c

@@ -85,6 +85,18 @@ static struct default_service default_services[] = {
 		.name			 = "corosync_pload",
 		.ver			 = 0,
 	},
+#ifdef HAVE_MONITORING
+	{
+		.name			 = "corosync_mon",
+		.ver			 = 0,
+	},
+#endif
+#ifdef HAVE_WATCHDOG
+	{
+		.name			 = "corosync_wd",
+		.ver			 = 0,
+	},
+#endif
 	{
 		.name			 = "corosync_quorum",
 		.ver			 = 0,

+ 3 - 1
include/corosync/corodefs.h

@@ -59,7 +59,9 @@ enum corosync_service_types {
 	NTF_SERVICE = 16,
 	AMF_V2_SERVICE = 17,
 	TST_SV1_SERVICE = 18,
-	TST_SV2_SERVICE = 19
+	TST_SV2_SERVICE = 19,
+	MON_SERVICE = 20,
+	WD_SERVICE = 21
 };
 
 #ifdef HAVE_SMALL_MEMORY_FOOTPRINT

+ 6 - 0
services/Makefile.am

@@ -38,6 +38,12 @@ INCLUDES		= -I$(top_builddir)/include -I$(top_srcdir)/include \
 			  -I$(top_srcdir)/include/corosync
 
 SERVICE_LCRSO		= evs cfg cpg confdb pload
+if BUILD_WATCHDOG
+SERVICE_LCRSO		+= wd
+endif
+if BUILD_MONITORING
+SERVICE_LCRSO		+= mon
+endif
 
 QUORUM_LCRSO		= votequorum testquorum
 

+ 635 - 0
services/mon.c

@@ -0,0 +1,635 @@
+/*
+ * Copyright (c) 2010 Red Hat, Inc.
+ *
+ * All rights reserved.
+ *
+ * Author: Angus Salkeld <asalkeld@redhat.com>
+ *
+ * This software licensed under BSD license, the text of which follows:
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the MontaVista Software, Inc. nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+ * THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <config.h>
+
+#include <unistd.h>
+#if defined(HAVE_LIBSTATGRAB)
+#include <statgrab.h>
+#endif
+
+#include <corosync/corotypes.h>
+#include <corosync/corodefs.h>
+#include <corosync/lcr/lcr_comp.h>
+#include <corosync/engine/coroapi.h>
+#include <corosync/list.h>
+#include <corosync/totem/coropoll.h>
+#include <corosync/engine/logsys.h>
+#include "../exec/fsm.h"
+
+
+LOGSYS_DECLARE_SUBSYS ("MON");
+
+#undef ENTER
+#define ENTER() log_printf (LOGSYS_LEVEL_INFO, "%s", __func__)
+
+/*
+ * Service Interfaces required by service_message_handler struct
+ */
+static int mon_exec_init_fn (
+	struct corosync_api_v1 *corosync_api);
+
+hdb_handle_t mon_poll = 0;
+static struct corosync_api_v1 *api;
+static hdb_handle_t resources_obj;
+static pthread_t mon_poll_thread;
+#define MON_DEFAULT_PERIOD 3
+
+struct corosync_service_engine mon_service_engine = {
+	.name			= "corosync resource monitoring service",
+	.id			= MON_SERVICE,
+	.priority		= 1,
+	.private_data_size	= 0,
+	.flow_control		= CS_LIB_FLOW_CONTROL_NOT_REQUIRED,
+	.lib_init_fn		= NULL,
+	.lib_exit_fn		= NULL,
+	.lib_engine		= NULL,
+	.lib_engine_count	= 0,
+	.exec_engine		= NULL,
+	.exec_engine_count	= 0,
+	.confchg_fn		= NULL,
+	.exec_init_fn		= mon_exec_init_fn,
+	.exec_dump_fn		= NULL,
+	.sync_mode		= CS_SYNC_V2
+};
+
+static DECLARE_LIST_INIT (confchg_notify);
+
+
+struct resource_instance {
+	hdb_handle_t handle;
+	const char *name;
+	poll_timer_handle timer_handle;
+	void (*update_stats_fn) (void *data);
+	struct cs_fsm fsm;
+	int32_t period;
+	objdb_value_types_t max_type;
+	union {
+		int32_t int32;
+		double dbl;
+	} max;
+};
+
+static void mem_update_stats_fn (void *data);
+static void load_update_stats_fn (void *data);
+
+static struct resource_instance memory_used_inst = {
+	.name = "memory_used",
+	.update_stats_fn = mem_update_stats_fn,
+	.max_type = OBJDB_VALUETYPE_INT32,
+	.max.int32 = INT32_MAX,
+	.period = MON_DEFAULT_PERIOD,
+};
+
+static struct resource_instance load_15min_inst = {
+	.name = "load_15min",
+	.update_stats_fn = load_update_stats_fn,
+	.max_type = OBJDB_VALUETYPE_DOUBLE,
+	.max.dbl = INT32_MAX,
+	.period = MON_DEFAULT_PERIOD,
+};
+
+
+/*
+ * F S M
+ */
+static void mon_config_changed (struct cs_fsm* fsm, int32_t event, void * data);
+static void mon_resource_failed (struct cs_fsm* fsm, int32_t event, void * data);
+
+const char * mon_ok_str = "ok";
+const char * mon_failed_str = "failed";
+const char * mon_failure_str = "failure";
+const char * mon_disabled_str = "disabled";
+const char * mon_config_changed_str = "config_changed";
+
+enum mon_resource_state {
+	MON_S_DISABLED,
+	MON_S_OK,
+	MON_S_FAILED
+};
+enum mon_resource_event {
+	MON_E_CONFIG_CHANGED,
+	MON_E_FAILURE
+};
+
+struct cs_fsm_entry mon_fsm_table[] = {
+	{ MON_S_DISABLED,	MON_E_CONFIG_CHANGED,	mon_config_changed,	{MON_S_DISABLED, MON_S_OK, -1} },
+	{ MON_S_DISABLED,	MON_E_FAILURE,		NULL,			{-1} },
+	{ MON_S_OK,		MON_E_CONFIG_CHANGED,	mon_config_changed,	{MON_S_OK, MON_S_DISABLED, -1} },
+	{ MON_S_OK,		MON_E_FAILURE,		mon_resource_failed,	{MON_S_FAILED, -1} },
+	{ MON_S_FAILED,		MON_E_CONFIG_CHANGED,	mon_config_changed,	{MON_S_OK, MON_S_DISABLED, -1} },
+	{ MON_S_FAILED,		MON_E_FAILURE,		NULL,			{-1} },
+};
+
+/*
+ * Dynamic loading descriptor
+ */
+
+static struct corosync_service_engine *mon_get_service_engine_ver0 (void);
+
+static struct corosync_service_engine_iface_ver0 mon_service_engine_iface = {
+	.corosync_get_service_engine_ver0	= mon_get_service_engine_ver0
+};
+
+static struct lcr_iface corosync_mon_ver0[1] = {
+	{
+		.name			= "corosync_mon",
+		.version		= 0,
+		.versions_replace	= 0,
+		.versions_replace_count = 0,
+		.dependencies		= 0,
+		.dependency_count	= 0,
+		.constructor		= NULL,
+		.destructor		= NULL,
+		.interfaces		= NULL,
+	}
+};
+
+static struct lcr_comp mon_comp_ver0 = {
+	.iface_count	= 1,
+	.ifaces		= corosync_mon_ver0
+};
+
+static struct corosync_service_engine *mon_get_service_engine_ver0 (void)
+{
+	return (&mon_service_engine);
+}
+
+#ifdef COROSYNC_SOLARIS
+void corosync_lcr_component_register (void);
+
+void corosync_lcr_component_register (void) {
+#else
+__attribute__ ((constructor)) static void corosync_lcr_component_register (void) {
+#endif
+	lcr_interfaces_set (&corosync_mon_ver0[0], &mon_service_engine_iface);
+
+	lcr_component_register (&mon_comp_ver0);
+}
+
+static const char * mon_res_state_to_str(struct cs_fsm* fsm,
+	int32_t state)
+{
+	switch (state) {
+	case MON_S_DISABLED:
+		return mon_disabled_str;
+		break;
+	case MON_S_OK:
+		return mon_ok_str;
+		break;
+	case MON_S_FAILED:
+		return mon_failed_str;
+		break;
+	}
+	return NULL;
+}
+
+static const char * mon_res_event_to_str(struct cs_fsm* fsm,
+	int32_t event)
+{
+	switch (event) {
+	case MON_E_CONFIG_CHANGED:
+		return mon_config_changed_str;
+		break;
+	case MON_E_FAILURE:
+		return mon_failure_str;
+		break;
+	}
+	return NULL;
+}
+
+static void mon_fsm_state_set (struct cs_fsm* fsm,
+	enum mon_resource_state next_state, struct resource_instance* inst)
+{
+	enum mon_resource_state prev_state = fsm->curr_state;
+	const char *state_str;
+
+	ENTER();
+
+	cs_fsm_state_set(fsm, next_state, inst);
+
+	if (prev_state == fsm->curr_state) {
+		return;
+	}
+	state_str = mon_res_state_to_str(fsm, fsm->curr_state);
+
+	api->object_key_replace (inst->handle,
+		"state", strlen ("state"),
+		state_str, strlen (state_str));
+}
+
+
+static void mon_config_changed (struct cs_fsm* fsm, int32_t event, void * data)
+{
+	struct resource_instance * inst = (struct resource_instance *)data;
+	char *str;
+	size_t str_len;
+	objdb_value_types_t type;
+	int32_t tmp_value;
+	int32_t res;
+
+	ENTER();
+
+	res = api->object_key_get_typed (inst->handle,
+			"poll_period",
+			(void**)&str, &str_len,
+			&type);
+	if (res == 0) {
+		tmp_value = strtol (str, NULL, 0);
+		if (tmp_value > 0 && tmp_value < 120) {
+			if (inst->period != tmp_value) {
+				inst->period = tmp_value;
+			}
+		}
+	}
+
+	res = api->object_key_get_typed (inst->handle, "max",
+			(void**)&str, &str_len, &type);
+	if (res != 0) {
+		if (inst->max_type == OBJDB_VALUETYPE_INT32) {
+			inst->max.int32 = INT32_MAX;
+		} else
+		if (inst->max_type == OBJDB_VALUETYPE_DOUBLE) {
+			inst->max.dbl = INT32_MAX;
+		}
+		mon_fsm_state_set (fsm, MON_S_DISABLED, inst);
+	} else {
+		if (inst->max_type == OBJDB_VALUETYPE_INT32) {
+			inst->max.int32 = strtol (str, NULL, 0);
+		} else
+		if (inst->max_type == OBJDB_VALUETYPE_DOUBLE) {
+			inst->max.dbl = strtod (str, NULL);
+		}
+		mon_fsm_state_set (fsm, MON_S_OK, inst);
+	}
+
+	if (mon_poll == 0) {
+		return;
+	}
+	poll_timer_delete (mon_poll, inst->timer_handle);
+	/*
+	 * run the updater, incase the period has shortened
+	 */
+	inst->update_stats_fn (inst);
+	poll_timer_add (mon_poll,
+		inst->period * 1000, NULL,
+		inst->update_stats_fn,
+		&inst->timer_handle);
+}
+
+void mon_resource_failed (struct cs_fsm* fsm, int32_t event, void * data)
+{
+	struct resource_instance * inst = (struct resource_instance *)data;
+	ENTER();
+	mon_fsm_state_set (fsm, MON_S_FAILED, inst);
+}
+
+static int32_t percent_mem_used_get(void)
+{
+#if defined(HAVE_LIBSTATGRAB)
+	sg_mem_stats *mem_stats;
+	sg_swap_stats *swap_stats;
+	long long total, freemem;
+
+	mem_stats = sg_get_mem_stats();
+	swap_stats = sg_get_swap_stats();
+
+	if (mem_stats == NULL || swap_stats != NULL) {
+		log_printf (LOGSYS_LEVEL_ERROR, "Unable to get memory stats: %s\n",
+			sg_str_error(sg_get_error()));
+		return -1;
+	}
+	total = mem_stats->total + swap_stats->total;
+	freemem = mem_stats->free + swap_stats->free;
+	return ((total - freemem) * 100) / total;
+#else
+#if defined(COROSYNC_LINUX)
+	char *line_ptr;
+	char line[512];
+	unsigned long long value;
+	FILE *f;
+	long long total = 0;
+	long long freemem = 0;
+
+	if ((f = fopen("/proc/meminfo", "r")) == NULL) {
+		return -1;
+	}
+
+	while ((line_ptr = fgets(line, sizeof(line), f)) != NULL) {
+		if (sscanf(line_ptr, "%*s %llu kB", &value) != 1) {
+			continue;
+		}
+		value *= 1024;
+
+		if (strncmp(line_ptr, "MemTotal:", 9) == 0) {
+			total += value;
+		} else if (strncmp(line_ptr, "MemFree:", 8) == 0) {
+			freemem += value;
+		} else if (strncmp(line_ptr, "SwapTotal:", 10) == 0) {
+			total += value;
+		} else if (strncmp(line_ptr, "SwapFree:", 9) == 0) {
+			freemem += value;
+		}
+	}
+
+	fclose(f);
+	return ((total - freemem) * 100) / total;
+#else
+#error need libstatgrab or linux.
+#endif /* COROSYNC_LINUX */
+#endif /* HAVE_LIBSTATGRAB */
+}
+
+
+static void mem_update_stats_fn (void *data)
+{
+	struct resource_instance * inst = (struct resource_instance *)data;
+	int32_t new_value;
+	uint64_t timestamp;
+
+	new_value = percent_mem_used_get();
+	if (new_value > 0) {
+		api->object_key_replace (inst->handle,
+			"current", strlen("current"),
+			&new_value, sizeof(new_value));
+
+		timestamp = time (NULL);
+
+		api->object_key_replace (inst->handle,
+			"last_updated", strlen("last_updated"),
+			&timestamp, sizeof(time_t));
+
+		if (new_value > inst->max.int32) {
+			cs_fsm_process (&inst->fsm, MON_E_FAILURE, inst);
+		}
+	}
+	poll_timer_add (mon_poll,
+		inst->period * 1000, inst,
+		inst->update_stats_fn,
+		&inst->timer_handle);
+}
+
+static double min15_loadavg_get(void)
+{
+#if defined(HAVE_LIBSTATGRAB)
+	sg_load_stats *load_stats;
+	load_stats = sg_get_load_stats ();
+	if (load_stats == NULL) {
+		log_printf (LOGSYS_LEVEL_ERROR, "Unable to get load stats: %s\n",
+			sg_str_error (sg_get_error()));
+		return -1;
+	}
+	return load_stats->min15;
+#else
+#if defined(COROSYNC_LINUX)
+	double loadav[3];
+	if (getloadavg(loadav,3) < 0) {
+		return -1;
+	}
+	return loadav[2];
+#else
+#error need libstatgrab or linux.
+#endif /* COROSYNC_LINUX */
+#endif /* HAVE_LIBSTATGRAB */
+}
+
+static void load_update_stats_fn (void *data)
+{
+	struct resource_instance * inst = (struct resource_instance *)data;
+	uint64_t timestamp;
+	int32_t res = 0;
+	double min15 = min15_loadavg_get();
+
+	if (min15 < 0) {
+	}
+	res = api->object_key_replace (inst->handle,
+		"current", strlen("current"),
+		&min15, sizeof (min15));
+	if (res != 0)
+		log_printf (LOGSYS_LEVEL_ERROR, "replace current failed: %d", res);
+
+	timestamp = cs_timestamp_get();
+
+	res = api->object_key_replace (inst->handle,
+		"last_updated", strlen("last_updated"),
+		&timestamp, sizeof(uint64_t));
+	if (res != 0)
+		log_printf (LOGSYS_LEVEL_ERROR, "replace last_updated failed: %d", res);
+
+	if (min15 > inst->max.dbl) {
+		cs_fsm_process (&inst->fsm, MON_E_FAILURE, &inst);
+	}
+
+	poll_timer_add (mon_poll,
+		inst->period * 1000, inst,
+		inst->update_stats_fn,
+		&inst->timer_handle);
+}
+
+static void *mon_thread_handler (void * unused)
+{
+#ifdef HAVE_LIBSTATGRAB
+	sg_init();
+#endif /* HAVE_LIBSTATGRAB */
+	mon_poll = poll_create ();
+
+	poll_timer_add (mon_poll,
+		memory_used_inst.period * 1000,
+		&memory_used_inst,
+		memory_used_inst.update_stats_fn,
+		&memory_used_inst.timer_handle);
+
+	poll_timer_add (mon_poll,
+		load_15min_inst.period * 1000,
+		&load_15min_inst,
+		load_15min_inst.update_stats_fn,
+		&load_15min_inst.timer_handle);
+	poll_run (mon_poll);
+
+	return NULL;
+}
+
+static int object_find_or_create (
+	hdb_handle_t parent_object_handle,
+	hdb_handle_t *object_handle,
+	const void *object_name,
+	size_t object_name_len)
+{
+	hdb_handle_t obj_finder;
+	hdb_handle_t obj;
+	int ret = -1;
+
+	api->object_find_create (
+		parent_object_handle,
+		object_name,
+		object_name_len,
+		&obj_finder);
+
+	if (api->object_find_next (obj_finder, &obj) == 0) {
+		/* found it */
+		*object_handle = obj;
+		ret = 0;
+	}
+	else {
+		ret = api->object_create (parent_object_handle,
+			object_handle,
+			object_name, object_name_len);
+	}
+
+	api->object_find_destroy (obj_finder);
+	return ret;
+}
+
+static void mon_key_change_notify (object_change_type_t change_type,
+	hdb_handle_t parent_object_handle,
+	hdb_handle_t object_handle,
+	const void *object_name_pt, size_t object_name_len,
+	const void *key_name_pt, size_t key_len,
+	const void *key_value_pt, size_t key_value_len,
+	void *priv_data_pt)
+{
+	struct resource_instance* inst = (struct resource_instance*)priv_data_pt;
+
+	if ((strcmp ((char*)key_name_pt, "max") == 0) ||
+		(strcmp ((char*)key_name_pt, "poll_period") == 0)) {
+		ENTER();
+		cs_fsm_process (&inst->fsm, MON_E_CONFIG_CHANGED, inst);
+	}
+}
+
+static void mon_instance_init (hdb_handle_t parent, struct resource_instance* inst)
+{
+	int32_t res;
+	char mon_period_str[32];
+	size_t mon_period_len;
+	objdb_value_types_t mon_period_type;
+	int32_t tmp_value;
+	int32_t zero_32 = 0;
+	time_t zero_64 = 0;
+	double zero_double = 0;
+
+	ENTER();
+
+	object_find_or_create (parent,
+		&inst->handle,
+		inst->name, strlen (inst->name));
+
+	if (inst->max_type == OBJDB_VALUETYPE_INT32) {
+		api->object_key_create_typed (inst->handle,
+			"current", &zero_32,
+			sizeof (zero_32), inst->max_type);
+	} else {
+		api->object_key_create_typed (inst->handle,
+			"current", &zero_double,
+			sizeof (zero_double), inst->max_type);
+	}
+
+	api->object_key_create_typed (inst->handle,
+		"last_updated", &zero_64,
+		sizeof (time_t), OBJDB_VALUETYPE_INT64);
+
+	api->object_key_create_typed (inst->handle,
+		"state", mon_disabled_str, strlen (mon_disabled_str),
+		OBJDB_VALUETYPE_STRING);
+
+	inst->fsm.name = inst->name;
+	inst->fsm.curr_entry = 0;
+	inst->fsm.curr_state = MON_S_DISABLED;
+	inst->fsm.table = mon_fsm_table;
+	inst->fsm.entries = sizeof(mon_fsm_table) / sizeof(struct cs_fsm_entry);
+	inst->fsm.state_to_str = mon_res_state_to_str;
+	inst->fsm.event_to_str = mon_res_event_to_str;
+
+	res = api->object_key_get_typed (inst->handle,
+			"poll_period",
+			(void**)&mon_period_str, &mon_period_len,
+			&mon_period_type);
+	if (res != 0) {
+		mon_period_len = snprintf (mon_period_str, 32, "%d",
+			inst->period);
+		api->object_key_create_typed (inst->handle,
+			"poll_period", &mon_period_str,
+			mon_period_len,
+			OBJDB_VALUETYPE_STRING);
+	}
+	else {
+		tmp_value = strtol (mon_period_str, NULL, 0);
+		if (tmp_value > 0 && tmp_value < 120)
+			inst->period = tmp_value;
+	}
+	cs_fsm_process (&inst->fsm, MON_E_CONFIG_CHANGED, inst);
+
+	poll_timer_add (mon_poll,
+		inst->period * 1000, inst,
+		inst->update_stats_fn,
+		&inst->timer_handle);
+
+	api->object_track_start (inst->handle, OBJECT_TRACK_DEPTH_ONE,
+		mon_key_change_notify,
+		NULL, NULL, NULL, NULL);
+
+}
+
+static int mon_exec_init_fn (
+	struct corosync_api_v1 *corosync_api)
+{
+	hdb_handle_t obj;
+	hdb_handle_t parent;
+
+#ifdef COROSYNC_SOLARIS
+	logsys_subsys_init();
+#endif
+	api = corosync_api;
+	ENTER();
+
+	object_find_or_create (OBJECT_PARENT_HANDLE,
+		&resources_obj,
+		"resources", strlen ("resources"));
+
+	object_find_or_create (resources_obj,
+		&obj,
+		"system", strlen ("system"));
+
+	parent = obj;
+
+	mon_instance_init (parent, &memory_used_inst);
+	mon_instance_init (parent, &load_15min_inst);
+
+
+	pthread_create (&mon_poll_thread, NULL, mon_thread_handler, NULL);
+
+	return 0;
+}
+
+

+ 755 - 0
services/wd.c

@@ -0,0 +1,755 @@
+/*
+ * Copyright (c) 2010 Red Hat, Inc.
+ *
+ * All rights reserved.
+ *
+ * Author: Angus Salkeld <asalkeld@redhat.com>
+ *
+ * This software licensed under BSD license, the text of which follows:
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the MontaVista Software, Inc. nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+ * THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <config.h>
+
+#include <unistd.h>
+#include <fcntl.h>
+#include <sys/ioctl.h>
+#include <linux/types.h>
+#include <linux/watchdog.h>
+#include <linux/reboot.h>
+
+#include <corosync/corotypes.h>
+#include <corosync/corodefs.h>
+#include <corosync/lcr/lcr_comp.h>
+#include <corosync/engine/coroapi.h>
+#include <corosync/list.h>
+#include <corosync/engine/logsys.h>
+#include "../exec/fsm.h"
+
+
+typedef enum {
+	WD_RESOURCE_GOOD,
+	WD_RESOURCE_FAILED,
+	WD_RESOURCE_STATE_UNKNOWN,
+	WD_RESOURCE_NOT_MONITORED
+} wd_resource_state_t;
+
+struct resource {
+	hdb_handle_t handle;
+	char *recovery;
+	char name[128];
+	time_t last_updated;
+	struct cs_fsm fsm;
+
+	corosync_timer_handle_t check_timer;
+	uint32_t check_timeout;
+};
+
+LOGSYS_DECLARE_SUBSYS("WD");
+
+/*
+ * Service Interfaces required by service_message_handler struct
+ */
+static int wd_exec_init_fn (
+	struct corosync_api_v1 *corosync_api);
+static int wd_exec_exit_fn (void);
+static void wd_resource_check_fn (void* resource_ref);
+
+static struct corosync_api_v1 *api;
+#define WD_DEFAULT_TIMEOUT 6
+static uint32_t watchdog_timeout = WD_DEFAULT_TIMEOUT;
+static uint32_t tickle_timeout = (WD_DEFAULT_TIMEOUT / 2);
+static int dog = -1;
+static corosync_timer_handle_t wd_timer;
+static hdb_handle_t resources_obj;
+static int watchdog_ok = 1;
+
+struct corosync_service_engine wd_service_engine = {
+	.name			= "corosync self-fencing service",
+	.id			= WD_SERVICE,
+	.priority		= 1,
+	.private_data_size	= 0,
+	.flow_control		= CS_LIB_FLOW_CONTROL_REQUIRED,
+	.lib_init_fn		= NULL,
+	.lib_exit_fn		= NULL,
+	.lib_engine		= NULL,
+	.lib_engine_count	= 0,
+	.exec_engine		= NULL,
+	.exec_engine_count	= 0,
+	.confchg_fn		= NULL,
+	.exec_init_fn		= wd_exec_init_fn,
+	.exec_exit_fn		= wd_exec_exit_fn,
+	.exec_dump_fn		= NULL,
+	.sync_mode		= CS_SYNC_V2
+};
+
+static DECLARE_LIST_INIT (confchg_notify);
+
+/*
+ * F S M
+ */
+static void wd_config_changed (struct cs_fsm* fsm, int32_t event, void * data);
+static void wd_resource_failed (struct cs_fsm* fsm, int32_t event, void * data);
+
+enum wd_resource_state {
+	WD_S_GOOD,
+	WD_S_FAILED,
+	WD_S_DISABLED
+};
+
+enum wd_resource_event {
+	WD_E_FAILURE,
+	WD_E_CONFIG_CHANGED
+};
+
+const char * wd_ok_str			= "ok";
+const char * wd_failed_str		= "failed";
+const char * wd_failure_str		= "failure";
+const char * wd_disabled_str		= "disabled";
+const char * wd_config_changed_str	= "config_changed";
+
+struct cs_fsm_entry wd_fsm_table[] = {
+	{ WD_S_DISABLED,	WD_E_CONFIG_CHANGED,	wd_config_changed,	{WD_S_DISABLED, WD_S_GOOD, -1} },
+	{ WD_S_DISABLED,	WD_E_FAILURE,		NULL,			{-1} },
+	{ WD_S_GOOD,		WD_E_CONFIG_CHANGED,	wd_config_changed,	{WD_S_GOOD, WD_S_DISABLED, -1} },
+	{ WD_S_GOOD,		WD_E_FAILURE,		wd_resource_failed,	{WD_S_FAILED, -1} },
+	{ WD_S_FAILED,		WD_E_CONFIG_CHANGED,	wd_config_changed,	{WD_S_GOOD, WD_S_DISABLED, -1} },
+	{ WD_S_FAILED,		WD_E_FAILURE,		NULL,			{-1} },
+};
+
+/*
+ * Dynamic loading descriptor
+ */
+
+static struct corosync_service_engine *wd_get_service_engine_ver0 (void);
+
+static struct corosync_service_engine_iface_ver0 wd_service_engine_iface = {
+	.corosync_get_service_engine_ver0	= wd_get_service_engine_ver0
+};
+
+static struct lcr_iface corosync_wd_ver0[1] = {
+	{
+		.name			= "corosync_wd",
+		.version		= 0,
+		.versions_replace	= 0,
+		.versions_replace_count = 0,
+		.dependencies		= 0,
+		.dependency_count	= 0,
+		.constructor		= NULL,
+		.destructor		= NULL,
+		.interfaces		= NULL,
+	}
+};
+
+static struct lcr_comp wd_comp_ver0 = {
+	.iface_count	= 1,
+	.ifaces		= corosync_wd_ver0
+};
+
+static struct corosync_service_engine *wd_get_service_engine_ver0 (void)
+{
+	return (&wd_service_engine);
+}
+
+#ifdef COROSYNC_SOLARIS
+void corosync_lcr_component_register (void);
+
+void corosync_lcr_component_register (void) {
+#else
+__attribute__ ((constructor)) static void corosync_lcr_component_register (void) {
+#endif
+	lcr_interfaces_set (&corosync_wd_ver0[0], &wd_service_engine_iface);
+
+	lcr_component_register (&wd_comp_ver0);
+}
+
+static int object_find_or_create (
+	hdb_handle_t parent_object_handle,
+	hdb_handle_t *object_handle,
+	const void *object_name,
+	size_t object_name_len)
+{
+	hdb_handle_t obj_finder;
+	hdb_handle_t obj;
+	int ret = -1;
+
+	api->object_find_create (
+		parent_object_handle,
+		object_name,
+		object_name_len,
+		&obj_finder);
+
+	if (api->object_find_next (obj_finder, &obj) == 0) {
+		/* found it */
+		*object_handle = obj;
+		ret = 0;
+	}
+	else {
+		ret = api->object_create (parent_object_handle,
+			object_handle,
+			object_name, object_name_len);
+	}
+
+	api->object_find_destroy (obj_finder);
+	return ret;
+}
+
+static const char * wd_res_state_to_str(struct cs_fsm* fsm,
+	int32_t state)
+{
+	switch (state) {
+	case WD_S_DISABLED:
+		return wd_disabled_str;
+		break;
+	case WD_S_GOOD:
+		return wd_ok_str;
+		break;
+	case WD_S_FAILED:
+		return wd_failed_str;
+		break;
+	}
+	return NULL;
+}
+
+static const char * wd_res_event_to_str(struct cs_fsm* fsm,
+	int32_t event)
+{
+	switch (event) {
+	case WD_E_CONFIG_CHANGED:
+		return wd_config_changed_str;
+		break;
+	case WD_E_FAILURE:
+		return wd_failure_str;
+		break;
+	}
+	return NULL;
+}
+
+/*
+ * returns (0 == OK, 1 == failed)
+ */
+static int32_t wd_resource_has_failed (struct resource *ref)
+{
+	hdb_handle_t resource = ref->handle;
+	int res;
+	char* state;
+	size_t state_len;
+	objdb_value_types_t type;
+	time_t *last_updated;
+	time_t my_time;
+	size_t last_updated_len;
+
+	res = api->object_key_get_typed (resource,
+		"last_updated", (void*)&last_updated, &last_updated_len, &type);
+	if (res != 0) {
+		/* key does not exist.
+		*/
+		return 1;
+	}
+	res = api->object_key_get_typed (resource,
+		"state", (void**)&state, &state_len, &type);
+	if (res != 0 ||	strncmp (state, "disabled", strlen ("disabled")) == 0) {
+		/* key does not exist.
+		*/
+		return 1;
+	}
+
+	my_time = time (NULL);
+
+	if ((*last_updated + ref->check_timeout) < my_time) {
+		log_printf (LOGSYS_LEVEL_INFO, "delayed %ld + %d < %ld",
+			*last_updated, ref->check_timeout, my_time);
+		return 1;
+	}
+
+	if ((*last_updated + ref->check_timeout) < my_time ||
+		strcmp (state, "bad") == 0) {
+		return 1;
+	}
+	return 0;
+}
+
+static void wd_config_changed (struct cs_fsm* fsm, int32_t event, void * data)
+{
+	int res;
+	size_t len;
+	char *state;
+	objdb_value_types_t type;
+	char mon_period_str[32];
+	int32_t tmp_value;
+	struct resource *ref = (struct resource*)data;
+
+	res = api->object_key_get_typed (ref->handle,
+			"poll_period",
+			(void**)&mon_period_str, &len,
+			&type);
+	if (res == 0) {
+		tmp_value = strtol (mon_period_str, NULL, 0);
+		if (tmp_value > 0 && tmp_value < 120)
+			ref->check_timeout = (tmp_value * 5)/4;
+	}
+
+	res = api->object_key_get_typed (ref->handle,
+		"recovery", (void*)&ref->recovery, &len, &type);
+	if (res != 0) {
+		/* key does not exist.
+		 */
+		log_printf (LOGSYS_LEVEL_WARNING,
+			"resource %s missing a recovery key.", ref->name);
+		cs_fsm_state_set(&ref->fsm, WD_S_DISABLED, ref);
+		return;
+	}
+	res = api->object_key_get_typed (ref->handle,
+		"state", (void*)&state, &len, &type);
+	if (res != 0) {
+		/* key does not exist.
+		*/
+		log_printf (LOGSYS_LEVEL_WARNING,
+			"resource %s missing a state key.", ref->name);
+		cs_fsm_state_set(&ref->fsm, WD_S_DISABLED, ref);
+		return;
+	}
+
+	cs_fsm_state_set(&ref->fsm, WD_S_GOOD, ref);
+
+	if (ref->check_timer) {
+		api->timer_delete(ref->check_timer);
+	}
+	api->timer_add_duration((unsigned long long)ref->check_timeout*1000000000,
+		ref,
+		wd_resource_check_fn, &ref->check_timer);
+
+}
+
+static void wd_resource_failed (struct cs_fsm* fsm, int32_t event, void * data)
+{
+	struct resource* ref = (struct resource*)data;
+
+	if (ref->check_timer) {
+		api->timer_delete(ref->check_timer);
+	}
+
+	log_printf (LOGSYS_LEVEL_CRIT, "%s resource \"%s\" failed!",
+		ref->recovery, (char*)ref->name);
+	if (strcmp (ref->recovery, "watchdog") == 0 ||
+	    strcmp (ref->recovery, "quit") == 0) {
+		watchdog_ok = 0;
+	}
+	else if (strcmp (ref->recovery, "reboot") == 0) {
+		//reboot(LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, LINUX_REBOOT_CMD_RESTART, NULL);
+	}
+	else if (strcmp (ref->recovery, "shutdown") == 0) {
+		//reboot(LINUX_REBOOT_MAGIC1, LINUX_REBOOT_MAGIC2, LINUX_REBOOT_CMD_POWER_OFF, NULL);
+	}
+	cs_fsm_state_set(fsm, WD_S_FAILED, data);
+}
+
+static void wd_key_changed(object_change_type_t change_type,
+	hdb_handle_t parent_object_handle,
+	hdb_handle_t object_handle,
+	const void *object_name_pt, size_t object_name_len,
+	const void *key_name_pt, size_t key_len,
+	const void *key_value_pt, size_t key_value_len,
+	void *priv_data_pt)
+{
+	struct resource* ref = (struct resource*)priv_data_pt;
+
+	if (strcmp(key_name_pt, "last_updated") == 0 ||
+		strcmp(key_name_pt, "current") == 0) {
+		return;
+	}
+//	log_printf (LOGSYS_LEVEL_WARNING,
+//		"watchdog resource key changed: %s.%s=%s ref=%p.",
+//		(char*)object_name_pt, (char*)key_name_pt, (char*)key_value_pt, ref);
+
+	if (ref == NULL) {
+		return;
+	}
+	cs_fsm_process(&ref->fsm, WD_E_CONFIG_CHANGED, ref);
+}
+
+static void wd_object_destroyed(
+	hdb_handle_t parent_object_handle,
+	const void *name_pt, size_t name_len,
+	void *priv_data_pt)
+{
+	struct resource* ref = (struct resource*)priv_data_pt;
+
+	log_printf (LOGSYS_LEVEL_WARNING,
+			"watchdog resource \"%s\" deleted from objdb!",
+			(char*)name_pt);
+
+	if (ref) {
+		api->timer_delete(ref->check_timer);
+		ref->check_timer = NULL;
+	}
+}
+
+static void wd_resource_check_fn (void* resource_ref)
+{
+	struct resource* ref = (struct resource*)resource_ref;
+
+	log_printf (LOGSYS_LEVEL_INFO,
+			"checking watchdog resource \"%s\".",
+			ref->name);
+	if (wd_resource_has_failed (ref) ) {
+		cs_fsm_process(&ref->fsm, WD_E_FAILURE, ref);
+		log_printf (LOGSYS_LEVEL_CRIT,
+			"watchdog resource \"%s\" failed!",
+			(char*)ref->name);
+		return;
+	}
+	api->timer_add_duration((unsigned long long)ref->check_timeout*1000000000,
+		ref, wd_resource_check_fn, &ref->check_timer);
+}
+
+
+static void wd_resource_create (hdb_handle_t resource_obj)
+{
+	int res;
+	size_t len;
+	char *state;
+	objdb_value_types_t type;
+	char mon_period_str[32];
+	int32_t tmp_value;
+	struct resource *ref = malloc (sizeof (struct resource));
+
+	ref->handle = resource_obj;
+	ref->check_timeout = WD_DEFAULT_TIMEOUT;
+	ref->check_timer = NULL;
+	api->object_name_get (resource_obj,
+		ref->name,
+		&len);
+	ref->name[len] = '\0';
+	ref->fsm.name = ref->name;
+	ref->fsm.table = wd_fsm_table;
+	ref->fsm.entries = sizeof(wd_fsm_table) / sizeof(struct cs_fsm_entry);
+	ref->fsm.curr_entry = 0;
+	ref->fsm.curr_state = WD_S_DISABLED;
+	ref->fsm.state_to_str = wd_res_state_to_str;
+	ref->fsm.event_to_str = wd_res_event_to_str;
+	api->object_priv_set (resource_obj, NULL);
+
+	res = api->object_key_get_typed (resource_obj,
+			"poll_period",
+			(void**)&mon_period_str, &len,
+			&type);
+	if (res != 0) {
+		log_printf (LOGSYS_LEVEL_ERROR, "%s : %d",__func__, res);
+		len = snprintf (mon_period_str, 32, "%d", ref->check_timeout);
+		api->object_key_create_typed (resource_obj,
+			"poll_period", &mon_period_str,
+			len,
+			OBJDB_VALUETYPE_STRING);
+	}
+	else {
+		tmp_value = strtol (mon_period_str, NULL, 0);
+		if (tmp_value > 0 && tmp_value < 120)
+			ref->check_timeout = (tmp_value * 5)/4;
+	}
+
+	api->object_track_start (resource_obj, OBJECT_TRACK_DEPTH_ONE,
+			wd_key_changed, NULL, wd_object_destroyed,
+			NULL, ref);
+
+	res = api->object_key_get_typed (resource_obj,
+		"recovery", (void*)&ref->recovery, &len, &type);
+	if (res != 0) {
+		/* key does not exist.
+		 */
+		log_printf (LOGSYS_LEVEL_WARNING,
+			"resource %s missing a recovery key.", ref->name);
+		return;
+	}
+	res = api->object_key_get_typed (resource_obj,
+		"state", (void*)&state, &len, &type);
+	if (res != 0) {
+		/* key does not exist.
+		*/
+		log_printf (LOGSYS_LEVEL_WARNING,
+			"resource %s missing a state key.", ref->name);
+		return;
+	}
+
+	res = api->object_key_get_typed (resource_obj,
+		"last_updated", (void*)&ref->last_updated, &len, &type);
+	if (res != 0) {
+		/* key does not exist.
+		 */
+		ref->last_updated = 0;
+	}
+
+	api->timer_add_duration((unsigned long long)ref->check_timeout*1000000000,
+		ref,
+		wd_resource_check_fn, &ref->check_timer);
+
+	cs_fsm_state_set(&ref->fsm, WD_S_GOOD, ref);
+}
+
+
+static void wd_tickle_fn (void* arg)
+{
+	ENTER();
+
+	if (watchdog_ok) {
+		if (dog > 0)
+			ioctl(dog, WDIOC_KEEPALIVE, &watchdog_ok);
+	}
+	else {
+		log_printf (LOGSYS_LEVEL_ALERT, "NOT tickling the watchdog!");
+	}
+
+	api->timer_add_duration((unsigned long long)tickle_timeout*1000000000, NULL,
+				wd_tickle_fn, &wd_timer);
+}
+
+static void wd_resource_object_created(hdb_handle_t parent_object_handle,
+	hdb_handle_t object_handle,
+	const void *name_pt, size_t name_len,
+	void *priv_data_pt)
+{
+	wd_resource_create (object_handle);
+}
+
+static void wd_scan_resources (void)
+{
+	hdb_handle_t obj_finder;
+	hdb_handle_t obj_finder2;
+	hdb_handle_t resource_type;
+	hdb_handle_t resource;
+	int res;
+
+	ENTER();
+
+	api->object_find_create (
+		OBJECT_PARENT_HANDLE,
+		"resources", strlen ("resources"),
+		&obj_finder);
+
+	res = api->object_find_next (obj_finder, &resources_obj);
+	api->object_find_destroy (obj_finder);
+	if (res != 0) {
+		log_printf (LOGSYS_LEVEL_INFO, "no resources.");
+		return;
+	}
+
+	/* this will be the system or process level
+	 */
+	api->object_find_create (
+		resources_obj,
+		NULL, 0,
+		&obj_finder);
+	while (api->object_find_next (obj_finder,
+			&resource_type) == 0) {
+
+		api->object_find_create (
+			resource_type,
+			NULL, 0,
+			&obj_finder2);
+
+		while (api->object_find_next (obj_finder2,
+				&resource) == 0) {
+
+			wd_resource_create (resource);
+		}
+		api->object_find_destroy (obj_finder2);
+
+		api->object_track_start (resource_type, OBJECT_TRACK_DEPTH_ONE,
+			NULL, wd_resource_object_created, NULL,
+			NULL, NULL);
+	}
+	api->object_find_destroy (obj_finder);
+}
+
+
+static void watchdog_timeout_apply (uint32_t new)
+{
+	struct watchdog_info ident;
+
+	if (new < 2) {
+		watchdog_timeout = 2;
+	}
+	else if (new > 120) {
+		watchdog_timeout = 120;
+	}
+	else {
+		watchdog_timeout = new;
+	}
+
+	if (dog > 0) {
+		ioctl(dog, WDIOC_GETSUPPORT, &ident);
+		if (ident.options & WDIOF_SETTIMEOUT) {
+			/* yay! the dog is trained.
+			 */
+			ioctl(dog, WDIOC_SETTIMEOUT, &watchdog_timeout);
+		}
+		ioctl(dog, WDIOC_GETTIMEOUT, &watchdog_timeout);
+	}
+	tickle_timeout = watchdog_timeout / 2;
+
+	log_printf (LOGSYS_LEVEL_DEBUG, "The Watchdog timeout is %d seconds\n", watchdog_timeout);
+	log_printf (LOGSYS_LEVEL_DEBUG, "The tickle timeout is %d seconds\n", tickle_timeout);
+}
+
+static int setup_watchdog(void)
+{
+	struct watchdog_info ident;
+
+	ENTER();
+	if (access ("/dev/watchdog", W_OK) != 0) {
+		log_printf (LOGSYS_LEVEL_WARNING, "No Watchdog, try modprobe <a watchdog>");
+		dog = -1;
+		return -1;
+	}
+
+	/* here goes, lets hope they have "Magic Close"
+	 */
+	dog = open("/dev/watchdog", O_WRONLY);
+
+	if (dog == -1) {
+		log_printf (LOGSYS_LEVEL_WARNING, "Watchdog exists but couldn't be opened.");
+		dog = -1;
+		return -1;
+	}
+
+	/* Right we have the dog.
+	 * Lets see what breed it is.
+	 */
+
+	ioctl(dog, WDIOC_GETSUPPORT, &ident);
+	log_printf (LOGSYS_LEVEL_INFO, "Watchdog is now been tickled by corosync.");
+	log_printf (LOGSYS_LEVEL_DEBUG, "%s", ident.identity);
+
+	watchdog_timeout_apply (watchdog_timeout);
+
+	ioctl(dog, WDIOC_SETOPTIONS, WDIOS_ENABLECARD);
+
+	return 0;
+}
+
+static void wd_top_level_key_changed(object_change_type_t change_type,
+	hdb_handle_t parent_object_handle,
+	hdb_handle_t object_handle,
+	const void *object_name_pt, size_t object_name_len,
+	const void *key_name_pt, size_t key_len,
+	const void *key_value_pt, size_t key_value_len,
+	void *priv_data_pt)
+{
+	uint32_t tmp_value;
+
+	ENTER();
+	if (change_type != OBJECT_KEY_DELETED &&
+		strncmp ((char*)key_name_pt, "watchdog_timeout", key_value_len) == 0) {
+		tmp_value = strtol (key_value_pt, NULL, 0);
+		watchdog_timeout_apply (tmp_value);
+	}
+	else {
+		watchdog_timeout_apply (WD_DEFAULT_TIMEOUT);
+	}
+	log_printf (LOGSYS_LEVEL_INFO, "new(%d) tickle_timeout: %d", change_type, tickle_timeout);
+}
+
+
+static void watchdog_timeout_get_initial (void)
+{
+	int32_t res;
+	char watchdog_timeout_str[32];
+	size_t watchdog_timeout_len;
+	objdb_value_types_t watchdog_timeout_type;
+	uint32_t tmp_value;
+
+	ENTER();
+
+	res = api->object_key_get_typed (resources_obj,
+			"watchdog_timeout",
+			(void**)&watchdog_timeout_str, &watchdog_timeout_len,
+			&watchdog_timeout_type);
+	if (res != 0) {
+		watchdog_timeout_apply (WD_DEFAULT_TIMEOUT);
+
+		watchdog_timeout_len = snprintf (watchdog_timeout_str, 32, "%d", watchdog_timeout);
+		api->object_key_create_typed (resources_obj,
+			"watchdog_timeout", &watchdog_timeout_str,
+			watchdog_timeout_len,
+			OBJDB_VALUETYPE_STRING);
+	}
+	else {
+		tmp_value = strtol (watchdog_timeout_str, NULL, 0);
+		watchdog_timeout_apply (tmp_value);
+	}
+
+	api->object_track_start (resources_obj, OBJECT_TRACK_DEPTH_ONE,
+		wd_top_level_key_changed, NULL, NULL,
+		NULL, NULL);
+
+}
+
+static int wd_exec_init_fn (
+	struct corosync_api_v1 *corosync_api)
+{
+	hdb_handle_t obj;
+
+	ENTER();
+#ifdef COROSYNC_SOLARIS
+	logsys_subsys_init();
+#endif
+	api = corosync_api;
+
+	object_find_or_create (OBJECT_PARENT_HANDLE,
+		&resources_obj,
+		"resources", strlen ("resources"));
+	object_find_or_create (resources_obj,
+		&obj,
+		"system", strlen ("system"));
+	object_find_or_create (resources_obj,
+		&obj,
+		"process", strlen ("process"));
+
+	watchdog_timeout_get_initial();
+
+	setup_watchdog();
+
+	wd_scan_resources();
+
+	api->timer_add_duration((unsigned long long)tickle_timeout*1000000000, NULL,
+				wd_tickle_fn, &wd_timer);
+
+	return 0;
+}
+
+static int wd_exec_exit_fn (void)
+{
+	char magic = 'V';
+	ENTER();
+
+	if (dog > 0) {
+		log_printf (LOGSYS_LEVEL_INFO, "magically closing the watchdog.");
+		write (dog, &magic, 1);
+	}
+	return 0;
+}
+
+