static gboolean remote_op_query_timeout(gpointer data)
{
remote_fencing_op_t *op = data;
op->query_timer = 0;
if(op->state == st_done) {
crm_debug("Operation %s for %s already completed", op->id, op->target);
} else if(op->state == st_exec) {
crm_debug("Operation %s for %s already in progress", op->id, op->target);
} else if(op->query_results) {
crm_debug("Query %s for %s complete: %d", op->id, op->target, op->state);
call_remote_stonith(op, NULL);
} else {
if(op->op_timer) {
g_source_remove(op->op_timer);
op->op_timer = 0;
}
remote_op_timeout(op);
}
return FALSE;
}
static gboolean
remote_op_timeout_one(gpointer userdata)
{
remote_fencing_op_t *op = userdata;
op->op_timer_one = 0;
crm_notice("Remote %s operation on %s for %s.%8s timed out",
op->action, op->target, op->client_name, op->id);
call_remote_stonith(op, NULL);
return FALSE;
}
int
process_remote_stonith_exec(xmlNode * msg)
{
int rc = 0;
const char *id = NULL;
const char *device = NULL;
remote_fencing_op_t *op = NULL;
xmlNode *dev = get_xpath_object("//@" F_STONITH_REMOTE_OP_ID, msg, LOG_ERR);
CRM_CHECK(dev != NULL, return -EPROTO);
id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID);
CRM_CHECK(id != NULL, return -EPROTO);
dev = get_xpath_object("//@" F_STONITH_RC, msg, LOG_ERR);
CRM_CHECK(dev != NULL, return -EPROTO);
crm_element_value_int(dev, F_STONITH_RC, &rc);
device = crm_element_value(dev, F_STONITH_DEVICE);
if (remote_op_list) {
op = g_hash_table_lookup(remote_op_list, id);
}
if (op == NULL && rc == pcmk_ok) {
/* Record successful fencing operations */
const char *client_id = crm_element_value(dev, F_STONITH_CLIENTID);
op = create_remote_stonith_op(client_id, dev, TRUE);
}
if (op == NULL) {
/* Could be for an event that began before we started */
/* TODO: Record the op for later querying */
crm_info("Unknown or expired remote op: %s", id);
return -EOPNOTSUPP;
}
if (op->devices && device && safe_str_neq(op->devices->data, device)) {
crm_err
("Received outdated reply for device %s (instead of %s) to %s node %s. Operation already timed out at remote level.",
device, op->devices->data, op->action, op->target);
return rc;
}
if (safe_str_eq(crm_element_value(msg, F_SUBTYPE), "broadcast")) {
crm_debug("Marking call to %s for %s on behalf of %s@%s.%.8s: %s (%d)",
op->action, op->target, op->client_name, op->id, op->originator,
pcmk_strerror(rc), rc);
if (rc == pcmk_ok) {
op->state = st_done;
} else {
op->state = st_failed;
}
remote_op_done(op, msg, rc, FALSE);
return pcmk_ok;
} else if (safe_str_neq(op->originator, stonith_our_uname)) {
/* If this isn't a remote level broadcast, and we are not the
* originator of the operation, we should not be receiving this msg. */
crm_err
("%s received non-broadcast fencing result for operation it does not own (device %s targeting %s)",
stonith_our_uname, device, op->target);
return rc;
}
if (is_set(op->call_options, st_opt_topology)) {
const char *device = crm_element_value(msg, F_STONITH_DEVICE);
crm_notice("Call to %s for %s on behalf of %s@%s: %s (%d)",
device, op->target, op->client_name, op->originator,
pcmk_strerror(rc), rc);
/* We own the op, and it is complete. broadcast the result to all nodes
* and notify our local clients. */
if (op->state == st_done) {
remote_op_done(op, msg, rc, FALSE);
return rc;
}
/* An operation completed succesfully but has not yet been marked as done.
* Continue the topology if more devices exist at the current level, otherwise
* mark as done. */
if (rc == pcmk_ok) {
if (op->devices) {
/* Success, are there any more? */
op->devices = op->devices->next;
}
/* if no more devices at this fencing level, we are done,
* else we need to contine with executing the next device in the list */
if (op->devices == NULL) {
crm_trace("Marking complex fencing op for %s as complete", op->target);
op->state = st_done;
remote_op_done(op, msg, rc, FALSE);
return rc;
}
} else {
/* This device failed, time to try another topology level. If no other
* levels are available, mark this operation as failed and report results. */
if (stonith_topology_next(op) != pcmk_ok) {
op->state = st_failed;
remote_op_done(op, msg, rc, FALSE);
return rc;
}
}
} else if (rc == pcmk_ok && op->devices == NULL) {
crm_trace("All done for %s", op->target);
op->state = st_done;
remote_op_done(op, msg, rc, FALSE);
return rc;
}
/* Retry on failure or execute the rest of the topology */
crm_trace("Next for %s on behalf of %s@%s (rc was %d)", op->target, op->originator,
op->client_name, rc);
call_remote_stonith(op, NULL);
return rc;
}
int
process_remote_stonith_query(xmlNode * msg)
{
int devices = 0;
gboolean host_is_target = FALSE;
const char *id = NULL;
const char *host = NULL;
remote_fencing_op_t *op = NULL;
st_query_result_t *result = NULL;
uint32_t active = fencing_active_peers();
xmlNode *dev = get_xpath_object("//@" F_STONITH_REMOTE_OP_ID, msg, LOG_ERR);
xmlNode *child = NULL;
CRM_CHECK(dev != NULL, return -EPROTO);
id = crm_element_value(dev, F_STONITH_REMOTE_OP_ID);
CRM_CHECK(id != NULL, return -EPROTO);
dev = get_xpath_object("//@st-available-devices", msg, LOG_ERR);
CRM_CHECK(dev != NULL, return -EPROTO);
crm_element_value_int(dev, "st-available-devices", &devices);
op = g_hash_table_lookup(remote_op_list, id);
if (op == NULL) {
crm_debug("Unknown or expired remote op: %s", id);
return -EOPNOTSUPP;
}
op->replies++;
host = crm_element_value(msg, F_ORIG);
host_is_target = safe_str_eq(host, op->target);
if (devices <= 0) {
/* If we're doing 'known' then we might need to fire anyway */
crm_trace("Query result from %s (%d devices)", host, devices);
if(op->state == st_query && (op->replies >= op->replies_expected || op->replies >= active)) {
crm_info("All queries have arrived, continuing (%d, %d, %d) ", op->replies_expected, active, op->replies);
call_remote_stonith(op, NULL);
}
return pcmk_ok;
} else if (host_is_target) {
if (op->call_options & st_opt_allow_suicide) {
crm_trace("Allowing %s to potentialy fence itself", op->target);
} else {
crm_info("Ignoring reply from %s, hosts are not permitted to commit suicide",
op->target);
return pcmk_ok;
}
}
crm_info("Query result %d of %d from %s (%d devices)", op->replies, op->replies_expected, host, devices);
result = calloc(1, sizeof(st_query_result_t));
result->host = strdup(host);
result->devices = devices;
result->custom_action_timeouts = g_hash_table_new_full(crm_str_hash, g_str_equal, free, NULL);
result->verified_devices = g_hash_table_new_full(crm_str_hash, g_str_equal, free, NULL);
for (child = __xml_first_child(dev); child != NULL; child = __xml_next(child)) {
const char *device = ID(child);
int action_timeout = 0;
int verified = 0;
if (device) {
result->device_list = g_list_prepend(result->device_list, strdup(device));
crm_element_value_int(child, F_STONITH_ACTION_TIMEOUT, &action_timeout);
crm_element_value_int(child, F_STONITH_DEVICE_VERIFIED, &verified);
if (action_timeout) {
crm_trace("Peer %s with device %s returned action timeout %d",
result->host, device, action_timeout);
g_hash_table_insert(result->custom_action_timeouts,
strdup(device), GINT_TO_POINTER(action_timeout));
}
if (verified) {
crm_trace("Peer %s has confirmed a verified device %s", result->host, device);
g_hash_table_insert(result->verified_devices,
strdup(device), GINT_TO_POINTER(verified));
}
}
}
CRM_CHECK(devices == g_list_length(result->device_list),
crm_err("Mis-match: Query claimed to have %d devices but %d found", devices,
g_list_length(result->device_list)));
op->query_results = g_list_insert_sorted(op->query_results, result, sort_peers);
if (is_set(op->call_options, st_opt_topology)) {
/* If we start the fencing before all the topology results are in,
* it is possible fencing levels will be skipped because of the missing
* query results. */
if (op->state == st_query && all_topology_devices_found(op)) {
/* All the query results are in for the topology, start the fencing ops. */
crm_trace("All topology devices found");
call_remote_stonith(op, result);
} else if(op->state == st_query && (op->replies >= op->replies_expected || op->replies >= active)) {
crm_info("All topology queries have arrived, continuing (%d, %d, %d) ", op->replies_expected, active, op->replies);
call_remote_stonith(op, NULL);
}
} else if (op->state == st_query) {
/* We have a result for a non-topology fencing op that looks promising,
* go ahead and start fencing before query timeout */
if (host_is_target == FALSE && g_hash_table_size(result->verified_devices)) {
/* we have a verified device living on a peer that is not the target */
crm_trace("Found %d verified devices", g_hash_table_size(result->verified_devices));
call_remote_stonith(op, result);
} else if (safe_str_eq(op->action, "on")) {
crm_trace("Unfencing %s", op->target);
call_remote_stonith(op, result);
} else if(op->replies >= op->replies_expected || op->replies >= active) {
crm_info("All queries have arrived, continuing (%d, %d, %d) ", op->replies_expected, active, op->replies);
call_remote_stonith(op, NULL);
} else {
crm_trace("Waiting for more peer results before launching fencing operation");
}
} else if (op->state == st_done) {
crm_info("Discarding query result from %s (%d devices): Operation is in state %d",
result->host, result->devices, op->state);
}
return pcmk_ok;
}
int process_remote_stonith_exec(xmlNode *msg)
{
int rc = 0;
const char *id = NULL;
remote_fencing_op_t *op = NULL;
xmlNode *dev = get_xpath_object("//@"F_STONITH_REMOTE, msg, LOG_ERR);
CRM_CHECK(dev != NULL, return -EPROTO);
id = crm_element_value(dev, F_STONITH_REMOTE);
CRM_CHECK(id != NULL, return -EPROTO);
dev = get_xpath_object("//@"F_STONITH_RC, msg, LOG_ERR);
CRM_CHECK(dev != NULL, return -EPROTO);
crm_element_value_int(dev, F_STONITH_RC, &rc);
if(remote_op_list) {
op = g_hash_table_lookup(remote_op_list, id);
}
if(op == NULL && rc == pcmk_ok) {
/* Record successful fencing operations */
const char *client_id = crm_element_value(msg, F_STONITH_CLIENTID);
op = create_remote_stonith_op(client_id, msg, TRUE);
}
if(op == NULL) {
/* Could be for an event that began before we started */
/* TODO: Record the op for later querying */
crm_info("Unknown or expired remote op: %s", id);
return -EOPNOTSUPP;
}
if(is_set(op->call_options, st_opt_topology)) {
const char *device = crm_element_value(msg, F_STONITH_DEVICE);
crm_notice("Call to %s for %s on behalf of %s: %s (%d)", device, op->target, op->originator, rc == pcmk_ok?"passed":"failed", rc);
if(safe_str_eq(op->originator, stonith_our_uname)) {
if(op->state == st_done) {
remote_op_done(op, msg, rc);
return rc;
} else if(rc == pcmk_ok && op->devices) {
/* Success, are there any more? */
op->devices = op->devices->next;
}
if(op->devices == NULL) {
crm_trace("Broadcasting completion of complex fencing op for %s", op->target);
send_cluster_message(NULL, crm_msg_stonith_ng, msg, FALSE);
op->state = st_done;
return rc;
}
} else {
op->state = st_done;
remote_op_done(op, msg, rc);
}
} else if(rc == pcmk_ok && op->devices == NULL) {
crm_trace("All done for %s", op->target);
op->state = st_done;
remote_op_done(op, msg, rc);
return rc;
}
/* Retry on failure or execute the rest of the topology */
crm_trace("Next for %s (rc was %d)", op->target, rc);
call_remote_stonith(op, NULL);
return rc;
}
int process_remote_stonith_query(xmlNode *msg)
{
int devices = 0;
const char *id = NULL;
const char *host = NULL;
remote_fencing_op_t *op = NULL;
st_query_result_t *result = NULL;
xmlNode *dev = get_xpath_object("//@"F_STONITH_REMOTE, msg, LOG_ERR);
xmlNode *child = NULL;
CRM_CHECK(dev != NULL, return -EPROTO);
id = crm_element_value(dev, F_STONITH_REMOTE);
CRM_CHECK(id != NULL, return -EPROTO);
dev = get_xpath_object("//@st-available-devices", msg, LOG_ERR);
CRM_CHECK(dev != NULL, return -EPROTO);
crm_element_value_int(dev, "st-available-devices", &devices);
op = g_hash_table_lookup(remote_op_list, id);
if(op == NULL) {
crm_debug("Unknown or expired remote op: %s", id);
return -EOPNOTSUPP;
}
op->replies++;
host = crm_element_value(msg, F_ORIG);
if(devices <= 0) {
/* If we're doing 'known' then we might need to fire anyway */
crm_trace("Query result from %s (%d devices)", host, devices);
return pcmk_ok;
} else if(op->call_options & st_opt_allow_suicide) {
crm_trace("Allowing %s to potentialy fence itself", op->target);
} else if(safe_str_eq(host, op->target)) {
crm_info("Ignoring reply from %s, hosts are not permitted to commit suicide", op->target);
return pcmk_ok;
}
crm_debug("Query result from %s (%d devices)", host, devices);
result = calloc(1, sizeof(st_query_result_t));
result->host = strdup(host);
result->devices = devices;
for (child = __xml_first_child(dev); child != NULL; child = __xml_next(child)) {
const char *device = ID(child);
if(device) {
result->device_list = g_list_prepend(result->device_list, strdup(device));
}
}
CRM_CHECK(devices == g_list_length(result->device_list),
crm_err("Mis-match: Query claimed to have %d devices but %d found", devices, g_list_length(result->device_list)));
op->query_results = g_list_insert_sorted(op->query_results, result, sort_peers);
if(op->state == st_query && is_set(op->call_options, st_opt_all_replies) == FALSE) {
call_remote_stonith(op, result);
} else if(op->state == st_done) {
crm_info("Discarding query result from %s (%d devices): Operation is in state %d",
result->host, result->devices, op->state);
}
return pcmk_ok;
}