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);
ref->check_timer = NULL;
}
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(RB_AUTOBOOT);
}
else if (strcmp (ref->recovery, "shutdown") == 0) {
reboot(RB_POWER_OFF);
}
cs_fsm_state_set(fsm, WD_S_FAILED, data);
}
/*
* return 0 - fully configured
* return -1 - partially configured
*/
static int32_t wd_resource_create (hdb_handle_t resource_obj)
{
int res;
size_t len;
char *state;
objdb_value_types_t type;
char period_str[32];
char str_copy[256];
char *str;
uint64_t tmp_value;
struct resource *ref = malloc (sizeof (struct resource));
ref->handle = resource_obj;
ref->check_timeout = WD_DEFAULT_TIMEOUT_MS;
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_STOPPED;
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**)&str, &len,
&type);
if (res != 0) {
len = snprintf (period_str, 32, "%"PRIu64"", ref->check_timeout);
api->object_key_create_typed (resource_obj,
"poll_period", &period_str,
len,
OBJDB_VALUETYPE_STRING);
}
else {
memcpy(str_copy, str, len);
str_copy[len] = '\0';
if (str_to_uint64_t(str_copy, &tmp_value, WD_MIN_TIMEOUT_MS, WD_MAX_TIMEOUT_MS) == CS_OK) {
ref->check_timeout = tmp_value;
} else {
log_printf (LOGSYS_LEVEL_WARNING,
"Could NOT use poll_period:%s ms for resource %s",
str, ref->name);
}
}
api->object_track_start (resource_obj, OBJECT_TRACK_DEPTH_RECURSIVE,
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 -1;
}
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 -1;
}
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;
}
/*
* delay the first check to give the monitor time to start working.
*/
tmp_value = CS_MAX(ref->check_timeout * 2, WD_DEFAULT_TIMEOUT_MS);
api->timer_add_duration(tmp_value * MILLI_2_NANO_SECONDS,
ref,
wd_resource_check_fn, &ref->check_timer);
cs_fsm_state_set(&ref->fsm, WD_S_RUNNING, ref);
return 0;
}
/*
* return 0 - fully configured
* return -1 - partially configured
*/
static int32_t wd_resource_create (char *res_path, char *res_name)
{
char *state;
uint64_t tmp_value;
struct resource *ref = calloc (1, sizeof (struct resource));
char key_name[ICMAP_KEYNAME_MAXLEN];
strcpy(ref->res_path, res_path);
ref->check_timeout = WD_DEFAULT_TIMEOUT_MS;
ref->check_timer = 0;
strcpy(ref->name, res_name);
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_STOPPED;
ref->fsm.state_to_str = wd_res_state_to_str;
ref->fsm.event_to_str = wd_res_event_to_str;
snprintf(key_name, ICMAP_KEYNAME_MAXLEN, "%s%s", res_path, "poll_period");
if (icmap_get_uint64(key_name, &tmp_value) != CS_OK) {
icmap_set_uint64(key_name, ref->check_timeout);
} else {
if (tmp_value >= WD_MIN_TIMEOUT_MS && tmp_value <= WD_MAX_TIMEOUT_MS) {
ref->check_timeout = tmp_value;
} else {
log_printf (LOGSYS_LEVEL_WARNING,
"Could NOT use poll_period:%"PRIu64" ms for resource %s",
tmp_value, ref->name);
}
}
icmap_track_add(res_path,
ICMAP_TRACK_ADD | ICMAP_TRACK_MODIFY | ICMAP_TRACK_DELETE | ICMAP_TRACK_PREFIX,
wd_key_changed,
ref, &ref->icmap_track);
snprintf(key_name, ICMAP_KEYNAME_MAXLEN, "%s%s", res_path, "recovery");
if (icmap_get_string(key_name, &ref->recovery) != CS_OK) {
/* key does not exist.
*/
log_printf (LOGSYS_LEVEL_WARNING,
"resource %s missing a recovery key.", ref->name);
return -1;
}
snprintf(key_name, ICMAP_KEYNAME_MAXLEN, "%s%s", res_path, "state");
if (icmap_get_string(key_name, &state) != CS_OK) {
/* key does not exist.
*/
log_printf (LOGSYS_LEVEL_WARNING,
"resource %s missing a state key.", ref->name);
return -1;
}
snprintf(key_name, ICMAP_KEYNAME_MAXLEN, "%s%s", res_path, "last_updated");
if (icmap_get_uint64(key_name, &tmp_value) != CS_OK) {
/* key does not exist.
*/
ref->last_updated = 0;
} else {
ref->last_updated = tmp_value;
}
/*
* delay the first check to give the monitor time to start working.
*/
tmp_value = CS_MAX(ref->check_timeout * 2, WD_DEFAULT_TIMEOUT_MS);
api->timer_add_duration(tmp_value * MILLI_2_NANO_SECONDS,
ref,
wd_resource_check_fn, &ref->check_timer);
cs_fsm_state_set(&ref->fsm, WD_S_RUNNING, ref);
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 *str;
uint64_t tmp_value;
uint64_t next_timeout;
struct resource *ref = (struct resource*)data;
char str_copy[256];
next_timeout = ref->check_timeout;
res = api->object_key_get_typed (ref->handle,
"poll_period",
(void**)&str, &len,
&type);
if (res == 0) {
memcpy(str_copy, str, len);
str_copy[len] = '\0';
if (str_to_uint64_t(str_copy, &tmp_value, WD_MIN_TIMEOUT_MS, WD_MAX_TIMEOUT_MS) == CS_OK) {
log_printf (LOGSYS_LEVEL_DEBUG,
"poll_period changing from:%"PRIu64" to %"PRIu64".",
ref->check_timeout, tmp_value);
/*
* To easy in the transition between poll_period's we are going
* to make the first timeout the bigger of the new and old value.
* This is to give the monitoring system time to adjust.
*/
next_timeout = CS_MAX(tmp_value, ref->check_timeout);
ref->check_timeout = tmp_value;
} else {
log_printf (LOGSYS_LEVEL_WARNING,
"Could NOT use poll_period:%s ms for resource %s",
str, ref->name);
}
}
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_STOPPED, 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_STOPPED, ref);
return;
}
if (ref->check_timer) {
api->timer_delete(ref->check_timer);
ref->check_timer = NULL;
}
if (strcmp(wd_stopped_str, state) == 0) {
cs_fsm_state_set(&ref->fsm, WD_S_STOPPED, ref);
} else {
api->timer_add_duration(next_timeout * MILLI_2_NANO_SECONDS,
ref, wd_resource_check_fn, &ref->check_timer);
cs_fsm_state_set(&ref->fsm, WD_S_RUNNING, ref);
}
}
static void wd_config_changed (struct cs_fsm* fsm, int32_t event, void * data)
{
char *state;
uint64_t tmp_value;
uint64_t next_timeout;
struct resource *ref = (struct resource*)data;
char key_name[ICMAP_KEYNAME_MAXLEN];
next_timeout = ref->check_timeout;
snprintf(key_name, ICMAP_KEYNAME_MAXLEN, "%s%s", ref->res_path, "poll_period");
if (icmap_get_uint64(ref->res_path, &tmp_value) == CS_OK) {
if (tmp_value >= WD_MIN_TIMEOUT_MS && tmp_value <= WD_MAX_TIMEOUT_MS) {
log_printf (LOGSYS_LEVEL_DEBUG,
"poll_period changing from:%"PRIu64" to %"PRIu64".",
ref->check_timeout, tmp_value);
/*
* To easy in the transition between poll_period's we are going
* to make the first timeout the bigger of the new and old value.
* This is to give the monitoring system time to adjust.
*/
next_timeout = CS_MAX(tmp_value, ref->check_timeout);
ref->check_timeout = tmp_value;
} else {
log_printf (LOGSYS_LEVEL_WARNING,
"Could NOT use poll_period:%"PRIu64" ms for resource %s",
tmp_value, ref->name);
}
}
snprintf(key_name, ICMAP_KEYNAME_MAXLEN, "%s%s", ref->res_path, "recovery");
if (icmap_get_string(key_name, &ref->recovery) != CS_OK) {
/* 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_STOPPED, ref);
return;
}
snprintf(key_name, ICMAP_KEYNAME_MAXLEN, "%s%s", ref->res_path, "state");
if (icmap_get_string(key_name, &state) != CS_OK) {
/* 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_STOPPED, ref);
return;
}
if (ref->check_timer) {
api->timer_delete(ref->check_timer);
ref->check_timer = 0;
}
if (strcmp(wd_stopped_str, state) == 0) {
cs_fsm_state_set(&ref->fsm, WD_S_STOPPED, ref);
} else {
api->timer_add_duration(next_timeout * MILLI_2_NANO_SECONDS,
ref, wd_resource_check_fn, &ref->check_timer);
cs_fsm_state_set(&ref->fsm, WD_S_RUNNING, ref);
}
free(state);
}
