Page MenuHomeClusterLabs Projects
Files F1842429

No OneTemporary
Actions

View Options

diff --git a/daemons/fenced/fenced_history.c b/daemons/fenced/fenced_history.c
index b65b64c7f1..cd08d7460d 100644
--- a/daemons/fenced/fenced_history.c
+++ b/daemons/fenced/fenced_history.c
@@ -1,482 +1,487 @@
/*
* Copyright 2009-2018 Andrew Beekhof <andrew@beekhof.net>
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/ipc.h>
#include <crm/common/ipcs.h>
#include <crm/cluster/internal.h>
#include <crm/stonith-ng.h>
#include <crm/fencing/internal.h>
#include <crm/common/xml.h>
#include <pacemaker-fenced.h>
#define MAX_STONITH_HISTORY 500
/*!
* \internal
* \brief Send a broadcast to all nodes to trigger cleanup or
* history synchronisation
*
* \param[in] history Optional history to be attached
* \param[in] callopts We control cleanup via a flag in the callopts
* \param[in] target Cleanup can be limited to certain fence-targets
*/
static void
stonith_send_broadcast_history(xmlNode *history,
int callopts,
const char *target)
{
xmlNode *bcast = create_xml_node(NULL, "stonith_command");
xmlNode *data = create_xml_node(NULL, __FUNCTION__);
if (target) {
crm_xml_add(data, F_STONITH_TARGET, target);
}
crm_xml_add(bcast, F_TYPE, T_STONITH_NG);
crm_xml_add(bcast, F_SUBTYPE, "broadcast");
crm_xml_add(bcast, F_STONITH_OPERATION, STONITH_OP_FENCE_HISTORY);
crm_xml_add_int(bcast, F_STONITH_CALLOPTS, callopts);
if (history) {
add_node_copy(data, history);
}
add_message_xml(bcast, F_STONITH_CALLDATA, data);
send_cluster_message(NULL, crm_msg_stonith_ng, bcast, FALSE);
free_xml(data);
free_xml(bcast);
}
static gboolean
stonith_remove_history_entry (gpointer key,
gpointer value,
gpointer user_data)
{
remote_fencing_op_t *op = value;
const char *target = (const char *) user_data;
if ((op->state == st_failed) || (op->state == st_done)) {
if ((target) && (strcmp(op->target, target) != 0)) {
return FALSE;
}
return TRUE;
}
return FALSE; /* don't clean pending operations */
}
/*!
* \internal
* \brief Send out a cleanup broadcast or do a local history-cleanup
*
* \param[in] target Cleanup can be limited to certain fence-targets
* \param[in] broadcast Send out a cleanup broadcast
*/
static void
stonith_fence_history_cleanup(const char *target,
gboolean broadcast)
{
if (broadcast) {
stonith_send_broadcast_history(NULL,
st_opt_cleanup | st_opt_discard_reply,
target);
/* we'll do the local clean when we receive back our own broadcast */
} else if (stonith_remote_op_list) {
g_hash_table_foreach_remove(stonith_remote_op_list,
stonith_remove_history_entry,
(gpointer) target);
do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL);
}
}
/* keeping the length of fence-history within bounds
* =================================================
*
* If things are really running wild a lot of fencing-attempts
* might fill up the hash-map, eventually using up a lot
* of memory and creating huge history-sync messages.
* Before the history being synced across nodes at least
* the reboot of a cluster-node helped keeping the
* history within bounds even though not in a reliable
* manner.
*
* stonith_remote_op_list isn't sorted for time-stamps
* thus it would be kind of expensive to delete e.g.
* the oldest entry if it would grow past MAX_STONITH_HISTORY
* entries.
* It is more efficient to purge MAX_STONITH_HISTORY/2
* entries whenever the list grows beyond MAX_STONITH_HISTORY.
* (sort for age + purge the MAX_STONITH_HISTORY/2 oldest)
* That done on a per-node-base might raise the
* probability of large syncs to occur.
* Things like introducing a broadcast to purge
* MAX_STONITH_HISTORY/2 entries or not sync above a certain
* threshold coming to mind ...
* Simplest thing though is to purge the full history
* throughout the cluster once MAX_STONITH_HISTORY is reached.
* On the other hand this leads to purging the history in
* situations where it would be handy to have it probably.
*/
static int
op_time_sort(const void *a_voidp, const void *b_voidp)
{
const remote_fencing_op_t **a = (const remote_fencing_op_t **) a_voidp;
const remote_fencing_op_t **b = (const remote_fencing_op_t **) b_voidp;
gboolean a_pending = ((*a)->state != st_failed) && ((*a)->state != st_done);
gboolean b_pending = ((*b)->state != st_failed) && ((*b)->state != st_done);
if (a_pending && b_pending) {
return 0;
} else if (a_pending) {
return -1;
} else if (b_pending) {
return 1;
} else if ((*b)->completed == (*a)->completed) {
return 0;
} else if ((*b)->completed > (*a)->completed) {
return 1;
}
return -1;
}
/*!
* \internal
* \brief Do a local history-trim to MAX_STONITH_HISTORY / 2 entries
* once over MAX_STONITH_HISTORY
*/
void
stonith_fence_history_trim(void)
{
guint num_ops;
if (!stonith_remote_op_list) {
return;
}
num_ops = g_hash_table_size(stonith_remote_op_list);
if (num_ops > MAX_STONITH_HISTORY) {
remote_fencing_op_t *ops[num_ops];
remote_fencing_op_t *op = NULL;
GHashTableIter iter;
int i;
crm_trace("Fencing History growing beyond limit of %d so purge "
"half of failed/successful attempts", MAX_STONITH_HISTORY);
/* write all ops into an array */
i = 0;
g_hash_table_iter_init(&iter, stonith_remote_op_list);
while (g_hash_table_iter_next(&iter, NULL, (void **)&op)) {
ops[i++] = op;
}
/* run quicksort over the array so that we get pending ops
* first and then sorted most recent to oldest
*/
qsort(ops, num_ops, sizeof(remote_fencing_op_t *), op_time_sort);
/* purgest oldest half of the history entries */
for (i = MAX_STONITH_HISTORY / 2; i < num_ops; i++) {
/* keep pending ops even if they shouldn't fill more than
* half of our buffer
*/
if ((ops[i]->state == st_failed) || (ops[i]->state == st_done)) {
g_hash_table_remove(stonith_remote_op_list, ops[i]->id);
}
}
/* we've just purged valid data from the list so there is no need
* to create a notification - if displayed it can stay
*/
}
}
/*!
* \internal
* \brief Convert xml fence-history to a hash-table like stonith_remote_op_list
*
* \param[in] history Fence-history in xml
*
* \return Fence-history as hash-table
*/
static GHashTable *
stonith_xml_history_to_list(xmlNode *history)
{
xmlNode *xml_op = NULL;
GHashTable *rv = NULL;
init_stonith_remote_op_hash_table(&rv);
CRM_LOG_ASSERT(rv != NULL);
for (xml_op = __xml_first_child(history); xml_op != NULL;
xml_op = __xml_next(xml_op)) {
remote_fencing_op_t *op = NULL;
char *id = crm_element_value_copy(xml_op, F_STONITH_REMOTE_OP_ID);
int completed, state;
if (!id) {
crm_warn("History to convert to hashtable has no id in entry");
continue;
}
crm_trace("Attaching op %s to hashtable", id);
op = calloc(1, sizeof(remote_fencing_op_t));
op->id = id;
op->target = crm_element_value_copy(xml_op, F_STONITH_TARGET);
op->action = crm_element_value_copy(xml_op, F_STONITH_ACTION);
op->originator = crm_element_value_copy(xml_op, F_STONITH_ORIGIN);
op->delegate = crm_element_value_copy(xml_op, F_STONITH_DELEGATE);
op->client_name = crm_element_value_copy(xml_op, F_STONITH_CLIENTNAME);
crm_element_value_int(xml_op, F_STONITH_DATE, &completed);
op->completed = (time_t) completed;
crm_element_value_int(xml_op, F_STONITH_STATE, &state);
op->state = (enum op_state) state;
g_hash_table_replace(rv, id, op);
CRM_LOG_ASSERT(g_hash_table_lookup(rv, id) != NULL);
}
return rv;
}
/*!
* \internal
* \brief Craft xml difference between local fence-history and a history
* coming from remote
*
* \param[in] remote_history Fence-history as hash-table (may be NULL)
* \param[in] add_id If crafting the answer for an API
* history-request there is no need for the id
* \param[in] target Optionally limit to certain fence-target
*
* \return The fence-history as xml
*/
static xmlNode *
stonith_local_history_diff(GHashTable *remote_history,
gboolean add_id,
const char *target)
{
xmlNode *history = NULL;
int cnt = 0;
if (stonith_remote_op_list) {
GHashTableIter iter;
remote_fencing_op_t *op = NULL;
history = create_xml_node(NULL, F_STONITH_HISTORY_LIST);
g_hash_table_iter_init(&iter, stonith_remote_op_list);
while (g_hash_table_iter_next(&iter, NULL, (void **)&op)) {
xmlNode *entry = NULL;
if (remote_history &&
g_hash_table_lookup(remote_history, op->id)) {
continue; /* skip entries broadcasted already */
}
if (target && strcmp(op->target, target) != 0) {
continue;
}
cnt++;
crm_trace("Attaching op %s", op->id);
entry = create_xml_node(history, STONITH_OP_EXEC);
if (add_id) {
crm_xml_add(entry, F_STONITH_REMOTE_OP_ID, op->id);
}
crm_xml_add(entry, F_STONITH_TARGET, op->target);
crm_xml_add(entry, F_STONITH_ACTION, op->action);
crm_xml_add(entry, F_STONITH_ORIGIN, op->originator);
crm_xml_add(entry, F_STONITH_DELEGATE, op->delegate);
crm_xml_add(entry, F_STONITH_CLIENTNAME, op->client_name);
crm_xml_add_int(entry, F_STONITH_DATE, op->completed);
crm_xml_add_int(entry, F_STONITH_STATE, op->state);
}
}
if (cnt == 0) {
free_xml(history);
return NULL;
} else {
return history;
}
}
/*!
* \internal
* \brief Merge fence-history coming from remote into local history
*
* \param[in] history Hash-table holding remote history to be merged in
*/
static void
stonith_merge_in_history_list(GHashTable *history)
{
GHashTableIter iter;
remote_fencing_op_t *op = NULL;
gboolean updated = FALSE;
if (!history) {
return;
}
init_stonith_remote_op_hash_table(&stonith_remote_op_list);
g_hash_table_iter_init(&iter, history);
while (g_hash_table_iter_next(&iter, NULL, (void **)&op)) {
remote_fencing_op_t *stored_op =
g_hash_table_lookup(stonith_remote_op_list, op->id);
if (stored_op) {
continue; // Skip existent (@TODO state-merging might be desirable)
}
updated = TRUE;
g_hash_table_iter_steal(&iter);
if ((op->state != st_failed) &&
(op->state != st_done) &&
safe_str_eq(op->originator, stonith_our_uname)) {
crm_warn("received pending action we are supposed to be the "
"owner but it's not in our records -> fail it");
op->state = st_failed;
op->completed = time(NULL);
/* use -EHOSTUNREACH to not introduce a new return-code that might
trigger unexpected results at other places and to prevent
remote_op_done from setting the delegate if not present
*/
stonith_bcast_result_to_peers(op, -EHOSTUNREACH);
}
g_hash_table_insert(stonith_remote_op_list, op->id, op);
/* we could trim the history here but if we bail
* out after trim we might miss more recent entries
* of those that might still be in the list
* if we don't bail out trimming once is more
* efficient and memory overhead is minimal as
* we are just moving pointers from one hash to
* another
*/
}
stonith_fence_history_trim();
if (updated) {
do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY, 0, NULL);
}
g_hash_table_destroy(history); /* remove what is left */
}
/*!
* \internal
* \brief Handle fence-history messages (either from API or coming in as
* broadcasts
*
* \param[in] msg Request message
* \param[in] output In case of a request from the API used to craft
* a reply from
* \param[in] remote_peer
* \param[in] options call-options from the request
*
* \return always success as there is actully nothing that can go really wrong
*/
int
stonith_fence_history(xmlNode *msg, xmlNode **output,
const char *remote_peer, int options)
{
int rc = 0;
const char *target = NULL;
xmlNode *dev = get_xpath_object("//@" F_STONITH_TARGET, msg, LOG_TRACE);
xmlNode *out_history = NULL;
if (dev) {
target = crm_element_value(dev, F_STONITH_TARGET);
if (target && (options & st_opt_cs_nodeid)) {
int nodeid = crm_atoi(target, NULL);
crm_node_t *node = crm_find_known_peer_full(nodeid, NULL, CRM_GET_PEER_ANY);
if (node) {
target = node->uname;
}
}
}
if (options & st_opt_cleanup) {
crm_trace("Cleaning up operations on %s in %p", target,
stonith_remote_op_list);
stonith_fence_history_cleanup(target,
crm_element_value(msg, F_STONITH_CALLID) != NULL);
} else if (options & st_opt_broadcast) {
+ /* there is no clear sign atm for when a history sync
+ is done so send a notification for anything
+ that smells like history-sync
+ */
+ do_stonith_notify(0, T_STONITH_NOTIFY_HISTORY_SYNCED, 0, NULL);
if (crm_element_value(msg, F_STONITH_CALLID)) {
/* this is coming from the stonith-API
*
* craft a broadcast with node's history
* so that every node can merge and broadcast
* what it has on top
*/
out_history = stonith_local_history_diff(NULL, TRUE, NULL);
crm_trace("Broadcasting history to peers");
stonith_send_broadcast_history(out_history,
st_opt_broadcast | st_opt_discard_reply,
NULL);
} else if (remote_peer &&
!safe_str_eq(remote_peer, stonith_our_uname)) {
xmlNode *history =
get_xpath_object("//" F_STONITH_HISTORY_LIST, msg, LOG_TRACE);
GHashTable *received_history =
history?stonith_xml_history_to_list(history):NULL;
/* either a broadcast created directly upon stonith-API request
* or a diff as response to such a thing
*
* in both cases it may have a history or not
* if we have differential data
* merge in what we've received and stop
* otherwise broadcast what we have on top
* marking as differential and merge in afterwards
*/
if (!history ||
!crm_is_true(crm_element_value(history,
F_STONITH_DIFFERENTIAL))) {
out_history =
stonith_local_history_diff(received_history, TRUE, NULL);
if (out_history) {
crm_trace("Broadcasting history-diff to peers");
crm_xml_add(out_history, F_STONITH_DIFFERENTIAL,
XML_BOOLEAN_TRUE);
stonith_send_broadcast_history(out_history,
st_opt_broadcast | st_opt_discard_reply,
NULL);
} else {
crm_trace("History-diff is empty - skip broadcast");
}
}
stonith_merge_in_history_list(received_history);
} else {
crm_trace("Skipping history-query-broadcast (%s%s)"
" we sent ourselves",
remote_peer?"remote-peer=":"local-ipc",
remote_peer?remote_peer:"");
}
} else {
/* plain history request */
crm_trace("Looking for operations on %s in %p", target,
stonith_remote_op_list);
*output = stonith_local_history_diff(NULL, FALSE, target);
}
free_xml(out_history);
return rc;
}
diff --git a/daemons/fenced/pacemaker-fenced.c b/daemons/fenced/pacemaker-fenced.c
index 7e9bb07666..7a87f93f7c 100644
--- a/daemons/fenced/pacemaker-fenced.c
+++ b/daemons/fenced/pacemaker-fenced.c
@@ -1,1518 +1,1521 @@
/*
* Copyright 2009-2019 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <sys/param.h>
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
#include <sys/utsname.h>
#include <stdlib.h>
#include <errno.h>
#include <fcntl.h>
#include <inttypes.h> /* U32T ~ PRIu32, X32T ~ PRIx32 */
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/ipc.h>
#include <crm/common/ipcs.h>
#include <crm/cluster/internal.h>
#include <crm/stonith-ng.h>
#include <crm/fencing/internal.h>
#include <crm/common/xml.h>
#include <crm/common/mainloop.h>
#include <crm/cib/internal.h>
#include <crm/pengine/status.h>
#include <pacemaker-internal.h>
#include <pacemaker-fenced.h>
char *stonith_our_uname = NULL;
char *stonith_our_uuid = NULL;
long stonith_watchdog_timeout_ms = 0;
static GMainLoop *mainloop = NULL;
gboolean stand_alone = FALSE;
static gboolean no_cib_connect = FALSE;
static gboolean stonith_shutdown_flag = FALSE;
static qb_ipcs_service_t *ipcs = NULL;
static xmlNode *local_cib = NULL;
static pe_working_set_t *fenced_data_set = NULL;
static cib_t *cib_api = NULL;
static void *cib_library = NULL;
static void stonith_shutdown(int nsig);
static void stonith_cleanup(void);
static int32_t
st_ipc_accept(qb_ipcs_connection_t * c, uid_t uid, gid_t gid)
{
if (stonith_shutdown_flag) {
crm_info("Ignoring new client [%d] during shutdown", crm_ipcs_client_pid(c));
return -EPERM;
}
if (crm_client_new(c, uid, gid) == NULL) {
return -EIO;
}
return 0;
}
static void
st_ipc_created(qb_ipcs_connection_t * c)
{
crm_trace("Connection created for %p", c);
}
/* Exit code means? */
static int32_t
st_ipc_dispatch(qb_ipcs_connection_t * qbc, void *data, size_t size)
{
uint32_t id = 0;
uint32_t flags = 0;
int call_options = 0;
xmlNode *request = NULL;
crm_client_t *c = crm_client_get(qbc);
const char *op = NULL;
if (c == NULL) {
crm_info("Invalid client: %p", qbc);
return 0;
}
request = crm_ipcs_recv(c, data, size, &id, &flags);
if (request == NULL) {
crm_ipcs_send_ack(c, id, flags, "nack", __FUNCTION__, __LINE__);
return 0;
}
op = crm_element_value(request, F_CRM_TASK);
if(safe_str_eq(op, CRM_OP_RM_NODE_CACHE)) {
crm_xml_add(request, F_TYPE, T_STONITH_NG);
crm_xml_add(request, F_STONITH_OPERATION, op);
crm_xml_add(request, F_STONITH_CLIENTID, c->id);
crm_xml_add(request, F_STONITH_CLIENTNAME, crm_client_name(c));
crm_xml_add(request, F_STONITH_CLIENTNODE, stonith_our_uname);
send_cluster_message(NULL, crm_msg_stonith_ng, request, FALSE);
free_xml(request);
return 0;
}
if (c->name == NULL) {
const char *value = crm_element_value(request, F_STONITH_CLIENTNAME);
if (value == NULL) {
value = "unknown";
}
c->name = crm_strdup_printf("%s.%u", value, c->pid);
}
crm_element_value_int(request, F_STONITH_CALLOPTS, &call_options);
crm_trace("Flags %" X32T "/%u for command %" U32T " from %s",
flags, call_options, id, crm_client_name(c));
if (is_set(call_options, st_opt_sync_call)) {
CRM_ASSERT(flags & crm_ipc_client_response);
CRM_LOG_ASSERT(c->request_id == 0); /* This means the client has two synchronous events in-flight */
c->request_id = id; /* Reply only to the last one */
}
crm_xml_add(request, F_STONITH_CLIENTID, c->id);
crm_xml_add(request, F_STONITH_CLIENTNAME, crm_client_name(c));
crm_xml_add(request, F_STONITH_CLIENTNODE, stonith_our_uname);
crm_log_xml_trace(request, "Client[inbound]");
stonith_command(c, id, flags, request, NULL);
free_xml(request);
return 0;
}
/* Error code means? */
static int32_t
st_ipc_closed(qb_ipcs_connection_t * c)
{
crm_client_t *client = crm_client_get(c);
if (client == NULL) {
return 0;
}
crm_trace("Connection %p closed", c);
crm_client_destroy(client);
/* 0 means: yes, go ahead and destroy the connection */
return 0;
}
static void
st_ipc_destroy(qb_ipcs_connection_t * c)
{
crm_trace("Connection %p destroyed", c);
st_ipc_closed(c);
}
static void
stonith_peer_callback(xmlNode * msg, void *private_data)
{
const char *remote_peer = crm_element_value(msg, F_ORIG);
const char *op = crm_element_value(msg, F_STONITH_OPERATION);
if (crm_str_eq(op, "poke", TRUE)) {
return;
}
crm_log_xml_trace(msg, "Peer[inbound]");
stonith_command(NULL, 0, 0, msg, remote_peer);
}
#if SUPPORT_COROSYNC
static void
stonith_peer_ais_callback(cpg_handle_t handle,
const struct cpg_name *groupName,
uint32_t nodeid, uint32_t pid, void *msg, size_t msg_len)
{
uint32_t kind = 0;
xmlNode *xml = NULL;
const char *from = NULL;
char *data = pcmk_message_common_cs(handle, nodeid, pid, msg, &kind, &from);
if(data == NULL) {
return;
}
if (kind == crm_class_cluster) {
xml = string2xml(data);
if (xml == NULL) {
crm_err("Invalid XML: '%.120s'", data);
free(data);
return;
}
crm_xml_add(xml, F_ORIG, from);
/* crm_xml_add_int(xml, F_SEQ, wrapper->id); */
stonith_peer_callback(xml, NULL);
}
free_xml(xml);
free(data);
return;
}
static void
stonith_peer_cs_destroy(gpointer user_data)
{
crm_crit("Lost connection to cluster layer, shutting down");
stonith_shutdown(0);
}
#endif
void
do_local_reply(xmlNode * notify_src, const char *client_id, gboolean sync_reply, gboolean from_peer)
{
/* send callback to originating child */
crm_client_t *client_obj = NULL;
int local_rc = pcmk_ok;
crm_trace("Sending response");
client_obj = crm_client_get_by_id(client_id);
crm_trace("Sending callback to request originator");
if (client_obj == NULL) {
local_rc = -1;
crm_trace("No client to sent the response to. F_STONITH_CLIENTID not set.");
} else {
int rid = 0;
if (sync_reply) {
CRM_LOG_ASSERT(client_obj->request_id);
rid = client_obj->request_id;
client_obj->request_id = 0;
crm_trace("Sending response %d to %s %s",
rid, client_obj->name, from_peer ? "(originator of delegated request)" : "");
} else {
crm_trace("Sending an event to %s %s",
client_obj->name, from_peer ? "(originator of delegated request)" : "");
}
local_rc = crm_ipcs_send(client_obj, rid, notify_src, sync_reply?crm_ipc_flags_none:crm_ipc_server_event);
}
if (local_rc < pcmk_ok && client_obj != NULL) {
crm_warn("%sSync reply to %s failed: %s",
sync_reply ? "" : "A-",
client_obj ? client_obj->name : "<unknown>", pcmk_strerror(local_rc));
}
}
long long
get_stonith_flag(const char *name)
{
if (safe_str_eq(name, T_STONITH_NOTIFY_FENCE)) {
return st_callback_notify_fence;
} else if (safe_str_eq(name, STONITH_OP_DEVICE_ADD)) {
return st_callback_device_add;
} else if (safe_str_eq(name, STONITH_OP_DEVICE_DEL)) {
return st_callback_device_del;
} else if (safe_str_eq(name, T_STONITH_NOTIFY_HISTORY)) {
return st_callback_notify_history;
+ } else if (safe_str_eq(name, T_STONITH_NOTIFY_HISTORY_SYNCED)) {
+ return st_callback_notify_history_synced;
+
}
return st_callback_unknown;
}
static void
stonith_notify_client(gpointer key, gpointer value, gpointer user_data)
{
xmlNode *update_msg = user_data;
crm_client_t *client = value;
const char *type = NULL;
CRM_CHECK(client != NULL, return);
CRM_CHECK(update_msg != NULL, return);
type = crm_element_value(update_msg, F_SUBTYPE);
CRM_CHECK(type != NULL, crm_log_xml_err(update_msg, "notify"); return);
if (client->ipcs == NULL) {
crm_trace("Skipping client with NULL channel");
return;
}
if (client->options & get_stonith_flag(type)) {
int rc = crm_ipcs_send(client, 0, update_msg, crm_ipc_server_event | crm_ipc_server_error);
if (rc <= 0) {
crm_warn("%s notification of client %s.%.6s failed: %s (%d)",
type, crm_client_name(client), client->id, pcmk_strerror(rc), rc);
} else {
crm_trace("Sent %s notification to client %s.%.6s", type, crm_client_name(client),
client->id);
}
}
}
void
do_stonith_async_timeout_update(const char *client_id, const char *call_id, int timeout)
{
crm_client_t *client = NULL;
xmlNode *notify_data = NULL;
if (!timeout || !call_id || !client_id) {
return;
}
client = crm_client_get_by_id(client_id);
if (!client) {
return;
}
notify_data = create_xml_node(NULL, T_STONITH_TIMEOUT_VALUE);
crm_xml_add(notify_data, F_TYPE, T_STONITH_TIMEOUT_VALUE);
crm_xml_add(notify_data, F_STONITH_CALLID, call_id);
crm_xml_add_int(notify_data, F_STONITH_TIMEOUT, timeout);
crm_trace("timeout update is %d for client %s and call id %s", timeout, client_id, call_id);
if (client) {
crm_ipcs_send(client, 0, notify_data, crm_ipc_server_event);
}
free_xml(notify_data);
}
void
do_stonith_notify(int options, const char *type, int result, xmlNode * data)
{
/* TODO: Standardize the contents of data */
xmlNode *update_msg = create_xml_node(NULL, "notify");
CRM_CHECK(type != NULL,;);
crm_xml_add(update_msg, F_TYPE, T_STONITH_NOTIFY);
crm_xml_add(update_msg, F_SUBTYPE, type);
crm_xml_add(update_msg, F_STONITH_OPERATION, type);
crm_xml_add_int(update_msg, F_STONITH_RC, result);
if (data != NULL) {
add_message_xml(update_msg, F_STONITH_CALLDATA, data);
}
crm_trace("Notifying clients");
g_hash_table_foreach(client_connections, stonith_notify_client, update_msg);
free_xml(update_msg);
crm_trace("Notify complete");
}
static void
do_stonith_notify_config(int options, const char *op, int rc,
const char *desc, int active)
{
xmlNode *notify_data = create_xml_node(NULL, op);
CRM_CHECK(notify_data != NULL, return);
crm_xml_add(notify_data, F_STONITH_DEVICE, desc);
crm_xml_add_int(notify_data, F_STONITH_ACTIVE, active);
do_stonith_notify(options, op, rc, notify_data);
free_xml(notify_data);
}
void
do_stonith_notify_device(int options, const char *op, int rc, const char *desc)
{
do_stonith_notify_config(options, op, rc, desc, g_hash_table_size(device_list));
}
void
do_stonith_notify_level(int options, const char *op, int rc, const char *desc)
{
do_stonith_notify_config(options, op, rc, desc, g_hash_table_size(topology));
}
static void
topology_remove_helper(const char *node, int level)
{
int rc;
char *desc = NULL;
xmlNode *data = create_xml_node(NULL, XML_TAG_FENCING_LEVEL);
crm_xml_add(data, F_STONITH_ORIGIN, __FUNCTION__);
crm_xml_add_int(data, XML_ATTR_STONITH_INDEX, level);
crm_xml_add(data, XML_ATTR_STONITH_TARGET, node);
rc = stonith_level_remove(data, &desc);
do_stonith_notify_level(0, STONITH_OP_LEVEL_DEL, rc, desc);
free_xml(data);
free(desc);
}
static void
remove_cib_device(xmlXPathObjectPtr xpathObj)
{
int max = numXpathResults(xpathObj), lpc = 0;
for (lpc = 0; lpc < max; lpc++) {
const char *rsc_id = NULL;
const char *standard = NULL;
xmlNode *match = getXpathResult(xpathObj, lpc);
CRM_LOG_ASSERT(match != NULL);
if(match != NULL) {
standard = crm_element_value(match, XML_AGENT_ATTR_CLASS);
}
if (safe_str_neq(standard, PCMK_RESOURCE_CLASS_STONITH)) {
continue;
}
rsc_id = crm_element_value(match, XML_ATTR_ID);
stonith_device_remove(rsc_id, TRUE);
}
}
static void
handle_topology_change(xmlNode *match, bool remove)
{
int rc;
char *desc = NULL;
CRM_CHECK(match != NULL, return);
crm_trace("Updating %s", ID(match));
if(remove) {
int index = 0;
char *key = stonith_level_key(match, -1);
crm_element_value_int(match, XML_ATTR_STONITH_INDEX, &index);
topology_remove_helper(key, index);
free(key);
}
rc = stonith_level_register(match, &desc);
do_stonith_notify_level(0, STONITH_OP_LEVEL_ADD, rc, desc);
free(desc);
}
static void
remove_fencing_topology(xmlXPathObjectPtr xpathObj)
{
int max = numXpathResults(xpathObj), lpc = 0;
for (lpc = 0; lpc < max; lpc++) {
xmlNode *match = getXpathResult(xpathObj, lpc);
CRM_LOG_ASSERT(match != NULL);
if (match && crm_element_value(match, XML_DIFF_MARKER)) {
/* Deletion */
int index = 0;
char *target = stonith_level_key(match, -1);
crm_element_value_int(match, XML_ATTR_STONITH_INDEX, &index);
if (target == NULL) {
crm_err("Invalid fencing target in element %s", ID(match));
} else if (index <= 0) {
crm_err("Invalid level for %s in element %s", target, ID(match));
} else {
topology_remove_helper(target, index);
}
/* } else { Deal with modifications during the 'addition' stage */
}
}
}
static void
register_fencing_topology(xmlXPathObjectPtr xpathObj)
{
int max = numXpathResults(xpathObj), lpc = 0;
for (lpc = 0; lpc < max; lpc++) {
xmlNode *match = getXpathResult(xpathObj, lpc);
handle_topology_change(match, TRUE);
}
}
/* Fencing
<diff crm_feature_set="3.0.6">
<diff-removed>
<fencing-topology>
<fencing-level id="f-p1.1" target="pcmk-1" index="1" devices="poison-pill" __crm_diff_marker__="removed:top"/>
<fencing-level id="f-p1.2" target="pcmk-1" index="2" devices="power" __crm_diff_marker__="removed:top"/>
<fencing-level devices="disk,network" id="f-p2.1"/>
</fencing-topology>
</diff-removed>
<diff-added>
<fencing-topology>
<fencing-level id="f-p.1" target="pcmk-1" index="1" devices="poison-pill" __crm_diff_marker__="added:top"/>
<fencing-level id="f-p2.1" target="pcmk-2" index="1" devices="disk,something"/>
<fencing-level id="f-p3.1" target="pcmk-2" index="2" devices="power" __crm_diff_marker__="added:top"/>
</fencing-topology>
</diff-added>
</diff>
*/
static void
fencing_topology_init()
{
xmlXPathObjectPtr xpathObj = NULL;
const char *xpath = "//" XML_TAG_FENCING_LEVEL;
crm_trace("Full topology refresh");
free_topology_list();
init_topology_list();
/* Grab everything */
xpathObj = xpath_search(local_cib, xpath);
register_fencing_topology(xpathObj);
freeXpathObject(xpathObj);
}
#define rsc_name(x) x->clone_name?x->clone_name:x->id
/*!
* \internal
* \brief Check whether our uname is in a resource's allowed node list
*
* \param[in] rsc Resource to check
*
* \return Pointer to node object if found, NULL otherwise
*/
static node_t *
our_node_allowed_for(resource_t *rsc)
{
GHashTableIter iter;
node_t *node = NULL;
if (rsc && stonith_our_uname) {
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
if (node && strcmp(node->details->uname, stonith_our_uname) == 0) {
break;
}
node = NULL;
}
}
return node;
}
/*!
* \internal
* \brief If a resource or any of its children are STONITH devices, update their
* definitions given a cluster working set.
*
* \param[in] rsc Resource to check
* \param[in] data_set Cluster working set with device information
*/
static void cib_device_update(resource_t *rsc, pe_working_set_t *data_set)
{
node_t *node = NULL;
const char *value = NULL;
const char *rclass = NULL;
node_t *parent = NULL;
gboolean remove = TRUE;
/* If this is a complex resource, check children rather than this resource itself.
* TODO: Mark each installed device and remove if untouched when this process finishes.
*/
if(rsc->children) {
GListPtr gIter = NULL;
for (gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
cib_device_update(gIter->data, data_set);
if(pe_rsc_is_clone(rsc)) {
crm_trace("Only processing one copy of the clone %s", rsc->id);
break;
}
}
return;
}
/* We only care about STONITH resources. */
rclass = crm_element_value(rsc->xml, XML_AGENT_ATTR_CLASS);
if (safe_str_neq(rclass, PCMK_RESOURCE_CLASS_STONITH)) {
return;
}
/* If this STONITH resource is disabled, just remove it. */
value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
if (safe_str_eq(value, RSC_STOPPED)) {
crm_info("Device %s has been disabled", rsc->id);
goto update_done;
}
/* Check whether our node is allowed for this resource (and its parent if in a group) */
node = our_node_allowed_for(rsc);
if (rsc->parent && (rsc->parent->variant == pe_group)) {
parent = our_node_allowed_for(rsc->parent);
}
if(node == NULL) {
/* Our node is disallowed, so remove the device */
GHashTableIter iter;
crm_info("Device %s has been disabled on %s: unknown", rsc->id, stonith_our_uname);
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
crm_trace("Available: %s = %d", node->details->uname, node->weight);
}
goto update_done;
} else if(node->weight < 0 || (parent && parent->weight < 0)) {
/* Our node (or its group) is disallowed by score, so remove the device */
char *score = score2char((node->weight < 0) ? node->weight : parent->weight);
crm_info("Device %s has been disabled on %s: score=%s", rsc->id, stonith_our_uname, score);
free(score);
goto update_done;
} else {
/* Our node is allowed, so update the device information */
int rc;
xmlNode *data;
GHashTableIter gIter;
stonith_key_value_t *params = NULL;
const char *name = NULL;
const char *agent = crm_element_value(rsc->xml, XML_EXPR_ATTR_TYPE);
const char *rsc_provides = NULL;
crm_debug("Device %s is allowed on %s: score=%d", rsc->id, stonith_our_uname, node->weight);
get_rsc_attributes(rsc->parameters, rsc, node, data_set);
get_meta_attributes(rsc->meta, rsc, node, data_set);
rsc_provides = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_PROVIDES);
g_hash_table_iter_init(&gIter, rsc->parameters);
while (g_hash_table_iter_next(&gIter, (gpointer *) & name, (gpointer *) & value)) {
if (!name || !value) {
continue;
}
params = stonith_key_value_add(params, name, value);
crm_trace(" %s=%s", name, value);
}
remove = FALSE;
data = create_device_registration_xml(rsc_name(rsc), st_namespace_any,
agent, params, rsc_provides);
stonith_key_value_freeall(params, 1, 1);
rc = stonith_device_register(data, NULL, TRUE);
CRM_ASSERT(rc == pcmk_ok);
free_xml(data);
}
update_done:
if(remove && g_hash_table_lookup(device_list, rsc_name(rsc))) {
stonith_device_remove(rsc_name(rsc), TRUE);
}
}
/*!
* \internal
* \brief Update all STONITH device definitions based on current CIB
*/
static void
cib_devices_update(void)
{
GListPtr gIter = NULL;
crm_info("Updating devices to version %s.%s.%s",
crm_element_value(local_cib, XML_ATTR_GENERATION_ADMIN),
crm_element_value(local_cib, XML_ATTR_GENERATION),
crm_element_value(local_cib, XML_ATTR_NUMUPDATES));
CRM_ASSERT(fenced_data_set != NULL);
fenced_data_set->input = local_cib;
fenced_data_set->now = crm_time_new(NULL);
fenced_data_set->flags |= pe_flag_quick_location;
fenced_data_set->localhost = stonith_our_uname;
cluster_status(fenced_data_set);
pcmk__schedule_actions(fenced_data_set, NULL, NULL);
for (gIter = fenced_data_set->resources; gIter != NULL; gIter = gIter->next) {
cib_device_update(gIter->data, fenced_data_set);
}
fenced_data_set->input = NULL; // Wasn't a copy, so don't let API free it
pe_reset_working_set(fenced_data_set);
}
static void
update_cib_stonith_devices_v2(const char *event, xmlNode * msg)
{
xmlNode *change = NULL;
char *reason = NULL;
bool needs_update = FALSE;
xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT);
for (change = __xml_first_child(patchset); change != NULL; change = __xml_next(change)) {
const char *op = crm_element_value(change, XML_DIFF_OP);
const char *xpath = crm_element_value(change, XML_DIFF_PATH);
const char *shortpath = NULL;
if ((op == NULL) ||
(strcmp(op, "move") == 0) ||
strstr(xpath, "/"XML_CIB_TAG_STATUS)) {
continue;
} else if (safe_str_eq(op, "delete") && strstr(xpath, "/"XML_CIB_TAG_RESOURCE)) {
const char *rsc_id = NULL;
char *search = NULL;
char *mutable = NULL;
if (strstr(xpath, XML_TAG_ATTR_SETS) ||
strstr(xpath, XML_TAG_META_SETS)) {
needs_update = TRUE;
reason = strdup("(meta) attribute deleted from resource");
break;
}
mutable = strdup(xpath);
rsc_id = strstr(mutable, "primitive[@id=\'");
if (rsc_id != NULL) {
rsc_id += strlen("primitive[@id=\'");
search = strchr(rsc_id, '\'');
}
if (search != NULL) {
*search = 0;
stonith_device_remove(rsc_id, TRUE);
} else {
crm_warn("Ignoring malformed CIB update (resource deletion)");
}
free(mutable);
} else if (strstr(xpath, "/"XML_CIB_TAG_RESOURCES) ||
strstr(xpath, "/"XML_CIB_TAG_CONSTRAINTS) ||
strstr(xpath, "/"XML_CIB_TAG_RSCCONFIG)) {
shortpath = strrchr(xpath, '/'); CRM_ASSERT(shortpath);
reason = crm_strdup_printf("%s %s", op, shortpath+1);
needs_update = TRUE;
break;
}
}
if(needs_update) {
crm_info("Updating device list from the cib: %s", reason);
cib_devices_update();
} else {
crm_trace("No updates for device list found in cib");
}
free(reason);
}
static void
update_cib_stonith_devices_v1(const char *event, xmlNode * msg)
{
const char *reason = "none";
gboolean needs_update = FALSE;
xmlXPathObjectPtr xpath_obj = NULL;
/* process new constraints */
xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_CONS_TAG_RSC_LOCATION);
if (numXpathResults(xpath_obj) > 0) {
int max = numXpathResults(xpath_obj), lpc = 0;
/* Safest and simplest to always recompute */
needs_update = TRUE;
reason = "new location constraint";
for (lpc = 0; lpc < max; lpc++) {
xmlNode *match = getXpathResult(xpath_obj, lpc);
crm_log_xml_trace(match, "new constraint");
}
}
freeXpathObject(xpath_obj);
/* process deletions */
xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_CIB_TAG_RESOURCE);
if (numXpathResults(xpath_obj) > 0) {
remove_cib_device(xpath_obj);
}
freeXpathObject(xpath_obj);
/* process additions */
xpath_obj = xpath_search(msg, "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_CIB_TAG_RESOURCE);
if (numXpathResults(xpath_obj) > 0) {
int max = numXpathResults(xpath_obj), lpc = 0;
for (lpc = 0; lpc < max; lpc++) {
const char *rsc_id = NULL;
const char *standard = NULL;
xmlNode *match = getXpathResult(xpath_obj, lpc);
rsc_id = crm_element_value(match, XML_ATTR_ID);
standard = crm_element_value(match, XML_AGENT_ATTR_CLASS);
if (safe_str_neq(standard, PCMK_RESOURCE_CLASS_STONITH)) {
continue;
}
crm_trace("Fencing resource %s was added or modified", rsc_id);
reason = "new resource";
needs_update = TRUE;
}
}
freeXpathObject(xpath_obj);
if(needs_update) {
crm_info("Updating device list from the cib: %s", reason);
cib_devices_update();
}
}
static void
update_cib_stonith_devices(const char *event, xmlNode * msg)
{
int format = 1;
xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT);
CRM_ASSERT(patchset);
crm_element_value_int(patchset, "format", &format);
switch(format) {
case 1:
update_cib_stonith_devices_v1(event, msg);
break;
case 2:
update_cib_stonith_devices_v2(event, msg);
break;
default:
crm_warn("Unknown patch format: %d", format);
}
}
/* Needs to hold node name + attribute name + attribute value + 75 */
#define XPATH_MAX 512
/*!
* \internal
* \brief Check whether a node has a specific attribute name/value
*
* \param[in] node Name of node to check
* \param[in] name Name of an attribute to look for
* \param[in] value The value the named attribute needs to be set to in order to be considered a match
*
* \return TRUE if the locally cached CIB has the specified node attribute
*/
gboolean
node_has_attr(const char *node, const char *name, const char *value)
{
char xpath[XPATH_MAX];
xmlNode *match;
int n;
CRM_CHECK(local_cib != NULL, return FALSE);
/* Search for the node's attributes in the CIB. While the schema allows
* multiple sets of instance attributes, and allows instance attributes to
* use id-ref to reference values elsewhere, that is intended for resources,
* so we ignore that here.
*/
n = snprintf(xpath, XPATH_MAX, "//" XML_CIB_TAG_NODES
"/" XML_CIB_TAG_NODE "[@uname='%s']/" XML_TAG_ATTR_SETS
"/" XML_CIB_TAG_NVPAIR "[@name='%s' and @value='%s']",
node, name, value);
match = get_xpath_object(xpath, local_cib, LOG_TRACE);
CRM_CHECK(n < XPATH_MAX, return FALSE);
return (match != NULL);
}
static void
update_fencing_topology(const char *event, xmlNode * msg)
{
int format = 1;
const char *xpath;
xmlXPathObjectPtr xpathObj = NULL;
xmlNode *patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT);
CRM_ASSERT(patchset);
crm_element_value_int(patchset, "format", &format);
if(format == 1) {
/* Process deletions (only) */
xpath = "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_REMOVED "//" XML_TAG_FENCING_LEVEL;
xpathObj = xpath_search(msg, xpath);
remove_fencing_topology(xpathObj);
freeXpathObject(xpathObj);
/* Process additions and changes */
xpath = "//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED "//" XML_TAG_FENCING_LEVEL;
xpathObj = xpath_search(msg, xpath);
register_fencing_topology(xpathObj);
freeXpathObject(xpathObj);
} else if(format == 2) {
xmlNode *change = NULL;
int add[] = { 0, 0, 0 };
int del[] = { 0, 0, 0 };
xml_patch_versions(patchset, add, del);
for (change = __xml_first_child(patchset); change != NULL; change = __xml_next(change)) {
const char *op = crm_element_value(change, XML_DIFF_OP);
const char *xpath = crm_element_value(change, XML_DIFF_PATH);
if(op == NULL) {
continue;
} else if(strstr(xpath, "/" XML_TAG_FENCING_LEVEL) != NULL) {
/* Change to a specific entry */
crm_trace("Handling %s operation %d.%d.%d for %s", op, add[0], add[1], add[2], xpath);
if(strcmp(op, "move") == 0) {
continue;
} else if(strcmp(op, "create") == 0) {
handle_topology_change(change->children, FALSE);
} else if(strcmp(op, "modify") == 0) {
xmlNode *match = first_named_child(change, XML_DIFF_RESULT);
if(match) {
handle_topology_change(match->children, TRUE);
}
} else if(strcmp(op, "delete") == 0) {
/* Nuclear option, all we have is the path and an id... not enough to remove a specific entry */
crm_info("Re-initializing fencing topology after %s operation %d.%d.%d for %s",
op, add[0], add[1], add[2], xpath);
fencing_topology_init();
return;
}
} else if (strstr(xpath, "/" XML_TAG_FENCING_TOPOLOGY) != NULL) {
/* Change to the topology in general */
crm_info("Re-initializing fencing topology after top-level %s operation %d.%d.%d for %s",
op, add[0], add[1], add[2], xpath);
fencing_topology_init();
return;
} else if (strstr(xpath, "/" XML_CIB_TAG_CONFIGURATION)) {
/* Changes to the whole config section, possibly including the topology as a whild */
if(first_named_child(change, XML_TAG_FENCING_TOPOLOGY) == NULL) {
crm_trace("Nothing for us in %s operation %d.%d.%d for %s.",
op, add[0], add[1], add[2], xpath);
} else if(strcmp(op, "delete") == 0 || strcmp(op, "create") == 0) {
crm_info("Re-initializing fencing topology after top-level %s operation %d.%d.%d for %s.",
op, add[0], add[1], add[2], xpath);
fencing_topology_init();
return;
}
} else {
crm_trace("Nothing for us in %s operation %d.%d.%d for %s",
op, add[0], add[1], add[2], xpath);
}
}
} else {
crm_warn("Unknown patch format: %d", format);
}
}
static bool have_cib_devices = FALSE;
static void
update_cib_cache_cb(const char *event, xmlNode * msg)
{
int rc = pcmk_ok;
xmlNode *stonith_enabled_xml = NULL;
xmlNode *stonith_watchdog_xml = NULL;
const char *stonith_enabled_s = NULL;
static gboolean stonith_enabled_saved = TRUE;
if(!have_cib_devices) {
crm_trace("Skipping updates until we get a full dump");
return;
} else if(msg == NULL) {
crm_trace("Missing %s update", event);
return;
}
/* Maintain a local copy of the CIB so that we have full access
* to device definitions, location constraints, and node attributes
*/
if (local_cib != NULL) {
int rc = pcmk_ok;
xmlNode *patchset = NULL;
crm_element_value_int(msg, F_CIB_RC, &rc);
if (rc != pcmk_ok) {
return;
}
patchset = get_message_xml(msg, F_CIB_UPDATE_RESULT);
xml_log_patchset(LOG_TRACE, "Config update", patchset);
rc = xml_apply_patchset(local_cib, patchset, TRUE);
switch (rc) {
case pcmk_ok:
case -pcmk_err_old_data:
break;
case -pcmk_err_diff_resync:
case -pcmk_err_diff_failed:
crm_notice("[%s] Patch aborted: %s (%d)", event, pcmk_strerror(rc), rc);
free_xml(local_cib);
local_cib = NULL;
break;
default:
crm_warn("[%s] ABORTED: %s (%d)", event, pcmk_strerror(rc), rc);
free_xml(local_cib);
local_cib = NULL;
}
}
if (local_cib == NULL) {
crm_trace("Re-requesting the full cib");
rc = cib_api->cmds->query(cib_api, NULL, &local_cib, cib_scope_local | cib_sync_call);
if(rc != pcmk_ok) {
crm_err("Couldn't retrieve the CIB: %s (%d)", pcmk_strerror(rc), rc);
return;
}
CRM_ASSERT(local_cib != NULL);
stonith_enabled_saved = FALSE; /* Trigger a full refresh below */
}
crm_peer_caches_refresh(local_cib);
stonith_enabled_xml = get_xpath_object("//nvpair[@name='stonith-enabled']", local_cib, LOG_TRACE);
if (stonith_enabled_xml) {
stonith_enabled_s = crm_element_value(stonith_enabled_xml, XML_NVPAIR_ATTR_VALUE);
}
if (stonith_enabled_s == NULL || crm_is_true(stonith_enabled_s)) {
long timeout_ms = 0;
const char *value = NULL;
stonith_watchdog_xml = get_xpath_object("//nvpair[@name='stonith-watchdog-timeout']", local_cib, LOG_TRACE);
if (stonith_watchdog_xml) {
value = crm_element_value(stonith_watchdog_xml, XML_NVPAIR_ATTR_VALUE);
}
if(value) {
timeout_ms = crm_get_msec(value);
}
if (timeout_ms < 0) {
timeout_ms = crm_auto_watchdog_timeout();
}
if(timeout_ms != stonith_watchdog_timeout_ms) {
crm_notice("New watchdog timeout %lds (was %lds)", timeout_ms/1000, stonith_watchdog_timeout_ms/1000);
stonith_watchdog_timeout_ms = timeout_ms;
}
} else {
stonith_watchdog_timeout_ms = 0;
}
if (stonith_enabled_s && crm_is_true(stonith_enabled_s) == FALSE) {
crm_trace("Ignoring cib updates while stonith is disabled");
stonith_enabled_saved = FALSE;
return;
} else if (stonith_enabled_saved == FALSE) {
crm_info("Updating stonith device and topology lists now that stonith is enabled");
stonith_enabled_saved = TRUE;
fencing_topology_init();
cib_devices_update();
} else {
update_fencing_topology(event, msg);
update_cib_stonith_devices(event, msg);
}
}
static void
init_cib_cache_cb(xmlNode * msg, int call_id, int rc, xmlNode * output, void *user_data)
{
crm_info("Updating device list from the cib: init");
have_cib_devices = TRUE;
local_cib = copy_xml(output);
crm_peer_caches_refresh(local_cib);
fencing_topology_init();
cib_devices_update();
}
static void
stonith_shutdown(int nsig)
{
stonith_shutdown_flag = TRUE;
crm_info("Terminating with %d clients",
crm_hash_table_size(client_connections));
if (mainloop != NULL && g_main_loop_is_running(mainloop)) {
g_main_loop_quit(mainloop);
} else {
stonith_cleanup();
crm_exit(CRM_EX_OK);
}
}
static void
cib_connection_destroy(gpointer user_data)
{
if (stonith_shutdown_flag) {
crm_info("Connection to the CIB manager closed");
return;
} else {
crm_crit("Lost connection to the CIB manager, shutting down");
}
if (cib_api) {
cib_api->cmds->signoff(cib_api);
}
stonith_shutdown(0);
}
static void
stonith_cleanup(void)
{
if (cib_api) {
cib_api->cmds->signoff(cib_api);
}
if (ipcs) {
qb_ipcs_destroy(ipcs);
}
crm_peer_destroy();
crm_client_cleanup();
free_stonith_remote_op_list();
free_topology_list();
free_device_list();
free_metadata_cache();
free(stonith_our_uname);
stonith_our_uname = NULL;
free_xml(local_cib);
local_cib = NULL;
}
/* *INDENT-OFF* */
static struct crm_option long_options[] = {
{"stand-alone", 0, 0, 's'},
{"stand-alone-w-cpg", 0, 0, 'c'},
{"logfile", 1, 0, 'l'},
{"verbose", 0, 0, 'V'},
{"version", 0, 0, '$'},
{"help", 0, 0, '?'},
{0, 0, 0, 0}
};
/* *INDENT-ON* */
static void
setup_cib(void)
{
int rc, retries = 0;
static cib_t *(*cib_new_fn) (void) = NULL;
if (cib_new_fn == NULL) {
cib_new_fn = find_library_function(&cib_library, CIB_LIBRARY, "cib_new", TRUE);
}
if (cib_new_fn != NULL) {
cib_api = (*cib_new_fn) ();
}
if (cib_api == NULL) {
crm_err("No connection to the CIB manager");
return;
}
do {
sleep(retries);
rc = cib_api->cmds->signon(cib_api, CRM_SYSTEM_STONITHD, cib_command);
} while (rc == -ENOTCONN && ++retries < 5);
if (rc != pcmk_ok) {
crm_err("Could not connect to the CIB manager: %s (%d)", pcmk_strerror(rc), rc);
} else if (pcmk_ok !=
cib_api->cmds->add_notify_callback(cib_api, T_CIB_DIFF_NOTIFY, update_cib_cache_cb)) {
crm_err("Could not set CIB notification callback");
} else {
rc = cib_api->cmds->query(cib_api, NULL, NULL, cib_scope_local);
cib_api->cmds->register_callback(cib_api, rc, 120, FALSE, NULL, "init_cib_cache_cb",
init_cib_cache_cb);
cib_api->cmds->set_connection_dnotify(cib_api, cib_connection_destroy);
crm_info("Watching for stonith topology changes");
}
}
struct qb_ipcs_service_handlers ipc_callbacks = {
.connection_accept = st_ipc_accept,
.connection_created = st_ipc_created,
.msg_process = st_ipc_dispatch,
.connection_closed = st_ipc_closed,
.connection_destroyed = st_ipc_destroy
};
/*!
* \internal
* \brief Callback for peer status changes
*
* \param[in] type What changed
* \param[in] node What peer had the change
* \param[in] data Previous value of what changed
*/
static void
st_peer_update_callback(enum crm_status_type type, crm_node_t * node, const void *data)
{
if ((type != crm_status_processes) && !is_set(node->flags, crm_remote_node)) {
/*
* This is a hack until we can send to a nodeid and/or we fix node name lookups
* These messages are ignored in stonith_peer_callback()
*/
xmlNode *query = create_xml_node(NULL, "stonith_command");
crm_xml_add(query, F_XML_TAGNAME, "stonith_command");
crm_xml_add(query, F_TYPE, T_STONITH_NG);
crm_xml_add(query, F_STONITH_OPERATION, "poke");
crm_debug("Broadcasting our uname because of node %u", node->id);
send_cluster_message(NULL, crm_msg_stonith_ng, query, FALSE);
free_xml(query);
}
}
int
main(int argc, char **argv)
{
int flag;
int lpc = 0;
int argerr = 0;
int option_index = 0;
crm_cluster_t cluster;
const char *actions[] = { "reboot", "off", "on", "list", "monitor", "status" };
crm_ipc_t *old_instance = NULL;
crm_log_preinit(NULL, argc, argv);
crm_set_options(NULL, "mode [options]", long_options,
"Provides a summary of cluster's current state."
"\n\nOutputs varying levels of detail in a number of different formats.\n");
while (1) {
flag = crm_get_option(argc, argv, &option_index);
if (flag == -1) {
break;
}
switch (flag) {
case 'V':
crm_bump_log_level(argc, argv);
break;
case 'l':
crm_add_logfile(optarg);
break;
case 's':
stand_alone = TRUE;
break;
case 'c':
stand_alone = FALSE;
no_cib_connect = TRUE;
break;
case '$':
case '?':
crm_help(flag, CRM_EX_OK);
break;
default:
++argerr;
break;
}
}
if (argc - optind == 1 && safe_str_eq("metadata", argv[optind])) {
printf("<?xml version=\"1.0\"?><!DOCTYPE resource-agent SYSTEM \"ra-api-1.dtd\">\n");
printf("<resource-agent name=\"pacemaker-fenced\">\n");
printf(" <version>1.0</version>\n");
printf(" <longdesc lang=\"en\">Instance attributes available for all \"stonith\"-class resources"
" and used by Pacemaker's fence daemon, formerly known as stonithd</longdesc>\n");
printf(" <shortdesc lang=\"en\">Instance attributes available for all \"stonith\"-class resources</shortdesc>\n");
printf(" <parameters>\n");
#if 0
// priority is not implemented yet
printf(" <parameter name=\"priority\" unique=\"0\">\n");
printf(" <shortdesc lang=\"en\">Devices that are not in a topology "
"are tried in order of highest to lowest integer priority</shortdesc>\n");
printf(" <content type=\"integer\" default=\"0\"/>\n");
printf(" </parameter>\n");
#endif
printf(" <parameter name=\"%s\" unique=\"0\">\n", STONITH_ATTR_HOSTARG);
printf
(" <shortdesc lang=\"en\">Advanced use only: An alternate parameter to supply instead of 'port'</shortdesc>\n");
printf
(" <longdesc lang=\"en\">Some devices do not support the standard 'port' parameter or may provide additional ones.\n"
"Use this to specify an alternate, device-specific, parameter that should indicate the machine to be fenced.\n"
"A value of 'none' can be used to tell the cluster not to supply any additional parameters.\n"
" </longdesc>\n");
printf(" <content type=\"string\" default=\"port\"/>\n");
printf(" </parameter>\n");
printf(" <parameter name=\"%s\" unique=\"0\">\n", STONITH_ATTR_HOSTMAP);
printf
(" <shortdesc lang=\"en\">A mapping of host names to ports numbers for devices that do not support host names.</shortdesc>\n");
printf
(" <longdesc lang=\"en\">Eg. node1:1;node2:2,3 would tell the cluster to use port 1 for node1 and ports 2 and 3 for node2</longdesc>\n");
printf(" <content type=\"string\" default=\"\"/>\n");
printf(" </parameter>\n");
printf(" <parameter name=\"%s\" unique=\"0\">\n", STONITH_ATTR_HOSTLIST);
printf
(" <shortdesc lang=\"en\">A list of machines controlled by this device (Optional unless %s=static-list).</shortdesc>\n",
STONITH_ATTR_HOSTCHECK);
printf(" <content type=\"string\" default=\"\"/>\n");
printf(" </parameter>\n");
printf(" <parameter name=\"%s\" unique=\"0\">\n", STONITH_ATTR_HOSTCHECK);
printf
(" <shortdesc lang=\"en\">How to determine which machines are controlled by the device.</shortdesc>\n");
printf(" <longdesc lang=\"en\">Allowed values: dynamic-list "
"(query the device via the 'list' command), static-list "
"(check the " STONITH_ATTR_HOSTLIST " attribute), status "
"(query the device via the 'status' command), none (assume "
"every device can fence every machine)</longdesc>\n");
printf(" <content type=\"string\" default=\"dynamic-list\"/>\n");
printf(" </parameter>\n");
printf(" <parameter name=\"%s\" unique=\"0\">\n", STONITH_ATTR_DELAY_MAX);
printf
(" <shortdesc lang=\"en\">Enable a random delay for stonith actions and specify the maximum of random delay.</shortdesc>\n");
printf
(" <longdesc lang=\"en\">This prevents double fencing when using slow devices such as sbd.\n"
"Use this to enable a random delay for stonith actions.\n"
"The overall delay is derived from this random delay value adding a static delay so that the sum is kept below the maximum delay.</longdesc>\n");
printf(" <content type=\"time\" default=\"0s\"/>\n");
printf(" </parameter>\n");
printf(" <parameter name=\"%s\" unique=\"0\">\n", STONITH_ATTR_DELAY_BASE);
printf
(" <shortdesc lang=\"en\">Enable a base delay for stonith actions and specify base delay value.</shortdesc>\n");
printf
(" <longdesc lang=\"en\">This prevents double fencing when different delays are configured on the nodes.\n"
"Use this to enable a static delay for stonith actions.\n"
"The overall delay is derived from a random delay value adding this static delay so that the sum is kept below the maximum delay.</longdesc>\n");
printf(" <content type=\"time\" default=\"0s\"/>\n");
printf(" </parameter>\n");
printf(" <parameter name=\"%s\" unique=\"0\">\n", STONITH_ATTR_ACTION_LIMIT);
printf
(" <shortdesc lang=\"en\">The maximum number of actions can be performed in parallel on this device</shortdesc>\n");
printf
(" <longdesc lang=\"en\">Cluster property concurrent-fencing=true needs to be configured first.\n"
"Then use this to specify the maximum number of actions can be performed in parallel on this device. -1 is unlimited.</longdesc>\n");
printf(" <content type=\"integer\" default=\"1\"/>\n");
printf(" </parameter>\n");
for (lpc = 0; lpc < DIMOF(actions); lpc++) {
printf(" <parameter name=\"pcmk_%s_action\" unique=\"0\">\n", actions[lpc]);
printf
(" <shortdesc lang=\"en\">Advanced use only: An alternate command to run instead of '%s'</shortdesc>\n",
actions[lpc]);
printf
(" <longdesc lang=\"en\">Some devices do not support the standard commands or may provide additional ones.\n"
"Use this to specify an alternate, device-specific, command that implements the '%s' action.</longdesc>\n",
actions[lpc]);
printf(" <content type=\"string\" default=\"%s\"/>\n", actions[lpc]);
printf(" </parameter>\n");
printf(" <parameter name=\"pcmk_%s_timeout\" unique=\"0\">\n", actions[lpc]);
printf
(" <shortdesc lang=\"en\">Advanced use only: Specify an alternate timeout to use for %s actions instead of stonith-timeout</shortdesc>\n",
actions[lpc]);
printf
(" <longdesc lang=\"en\">Some devices need much more/less time to complete than normal.\n"
"Use this to specify an alternate, device-specific, timeout for '%s' actions.</longdesc>\n",
actions[lpc]);
printf(" <content type=\"time\" default=\"60s\"/>\n");
printf(" </parameter>\n");
printf(" <parameter name=\"pcmk_%s_retries\" unique=\"0\">\n", actions[lpc]);
printf
(" <shortdesc lang=\"en\">Advanced use only: The maximum number of times to retry the '%s' command within the timeout period</shortdesc>\n",
actions[lpc]);
printf(" <longdesc lang=\"en\">Some devices do not support multiple connections."
" Operations may 'fail' if the device is busy with another task so Pacemaker will automatically retry the operation, if there is time remaining."
" Use this option to alter the number of times Pacemaker retries '%s' actions before giving up."
"</longdesc>\n", actions[lpc]);
printf(" <content type=\"integer\" default=\"2\"/>\n");
printf(" </parameter>\n");
}
printf(" </parameters>\n");
printf("</resource-agent>\n");
return CRM_EX_OK;
}
if (optind != argc) {
++argerr;
}
if (argerr) {
crm_help('?', CRM_EX_USAGE);
}
crm_log_init(NULL, LOG_INFO, TRUE, FALSE, argc, argv, FALSE);
crm_notice("Starting Pacemaker fencer");
old_instance = crm_ipc_new("stonith-ng", 0);
if (crm_ipc_connect(old_instance)) {
/* IPC end-point already up */
crm_ipc_close(old_instance);
crm_ipc_destroy(old_instance);
crm_err("pacemaker-fenced is already active, aborting startup");
crm_exit(CRM_EX_OK);
} else {
/* not up or not authentic, we'll proceed either way */
crm_ipc_destroy(old_instance);
old_instance = NULL;
}
mainloop_add_signal(SIGTERM, stonith_shutdown);
crm_peer_init();
fenced_data_set = pe_new_working_set();
CRM_ASSERT(fenced_data_set != NULL);
if (stand_alone == FALSE) {
if (is_corosync_cluster()) {
#if SUPPORT_COROSYNC
cluster.destroy = stonith_peer_cs_destroy;
cluster.cpg.cpg_deliver_fn = stonith_peer_ais_callback;
cluster.cpg.cpg_confchg_fn = pcmk_cpg_membership;
#endif
}
crm_set_status_callback(&st_peer_update_callback);
if (crm_cluster_connect(&cluster) == FALSE) {
crm_crit("Cannot sign in to the cluster... terminating");
crm_exit(CRM_EX_FATAL);
}
stonith_our_uname = cluster.uname;
stonith_our_uuid = cluster.uuid;
if (no_cib_connect == FALSE) {
setup_cib();
}
} else {
stonith_our_uname = strdup("localhost");
}
init_device_list();
init_topology_list();
if(stonith_watchdog_timeout_ms > 0) {
int rc;
xmlNode *xml;
stonith_key_value_t *params = NULL;
params = stonith_key_value_add(params, STONITH_ATTR_HOSTLIST, stonith_our_uname);
xml = create_device_registration_xml("watchdog", st_namespace_internal,
STONITH_WATCHDOG_AGENT, params,
NULL);
stonith_key_value_freeall(params, 1, 1);
rc = stonith_device_register(xml, NULL, FALSE);
free_xml(xml);
if (rc != pcmk_ok) {
crm_crit("Cannot register watchdog pseudo fence agent");
crm_exit(CRM_EX_FATAL);
}
}
stonith_ipc_server_init(&ipcs, &ipc_callbacks);
/* Create the mainloop and run it... */
mainloop = g_main_loop_new(NULL, FALSE);
crm_notice("Pacemaker fencer successfully started and accepting connections");
g_main_loop_run(mainloop);
stonith_cleanup();
pe_free_working_set(fenced_data_set);
crm_exit(CRM_EX_OK);
}
diff --git a/daemons/fenced/pacemaker-fenced.h b/daemons/fenced/pacemaker-fenced.h
index a8531a69cd..583cb47a14 100644
--- a/daemons/fenced/pacemaker-fenced.h
+++ b/daemons/fenced/pacemaker-fenced.h
@@ -1,262 +1,263 @@
/*
* Copyright 2009-2019 the Pacemaker project contributors
*
* This source code is licensed under the GNU General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm/common/mainloop.h>
/*!
* \internal
* \brief Check to see if target was fenced in the last few seconds.
* \param tolerance, The number of seconds to look back in time
* \param target, The node to search for
* \param action, The action we want to match.
*
* \retval FALSE, not match
* \retval TRUE, fencing operation took place in the last 'tolerance' number of seconds.
*/
gboolean stonith_check_fence_tolerance(int tolerance, const char *target, const char *action);
enum st_device_flags
{
st_device_supports_list = 0x0001,
st_device_supports_status = 0x0002,
st_device_supports_reboot = 0x0004,
};
typedef struct stonith_device_s {
char *id;
char *agent;
char *namespace;
/*! list of actions that must execute on the target node. Used for unfencing */
char *on_target_actions;
GListPtr targets;
time_t targets_age;
gboolean has_attr_map;
/* should nodeid parameter for victim be included in agent arguments */
gboolean include_nodeid;
/* whether the cluster should automatically unfence nodes with the device */
gboolean automatic_unfencing;
guint priority;
enum st_device_flags flags;
GHashTable *params;
GHashTable *aliases;
GList *pending_ops;
crm_trigger_t *work;
xmlNode *agent_metadata;
/*! A verified device is one that has contacted the
* agent successfully to perform a monitor operation */
gboolean verified;
gboolean cib_registered;
gboolean api_registered;
} stonith_device_t;
/* These values are used to index certain arrays by "phase". Usually an
* operation has only one "phase", so phase is always zero. However, some
* reboots are remapped to "off" then "on", in which case "reboot" will be
* phase 0, "off" will be phase 1 and "on" will be phase 2.
*/
enum st_remap_phase {
st_phase_requested = 0,
st_phase_off = 1,
st_phase_on = 2,
st_phase_max = 3
};
typedef struct remote_fencing_op_s {
/* The unique id associated with this operation */
char *id;
/*! The node this operation will fence */
char *target;
/*! The fencing action to perform on the target. (reboot, on, off) */
char *action;
/*! When was the fencing action recorded (seconds since epoch) */
time_t created;
/*! Marks if the final notifications have been sent to local stonith clients. */
gboolean notify_sent;
/*! The number of query replies received */
guint replies;
/*! The number of query replies expected */
guint replies_expected;
/*! Does this node own control of this operation */
gboolean owner;
/*! After query is complete, This the high level timer that expires the entire operation */
guint op_timer_total;
/*! This timer expires the current fencing request. Many fencing
* requests may exist in a single operation */
guint op_timer_one;
/*! This timer expires the query request sent out to determine
* what nodes are contain what devices, and who those devices can fence */
guint query_timer;
/*! This is the default timeout to use for each fencing device if no
* custom timeout is received in the query. */
gint base_timeout;
/*! This is the calculated total timeout an operation can take before
* expiring. This is calculated by adding together all the timeout
* values associated with the devices this fencing operation may call */
gint total_timeout;
/*! Delegate is the node being asked to perform a fencing action
* on behalf of the node that owns the remote operation. Some operations
* will involve multiple delegates. This value represents the final delegate
* that is used. */
char *delegate;
/*! The point at which the remote operation completed */
time_t completed;
/*! The stonith_call_options associated with this remote operation */
long long call_options;
/*! The current state of the remote operation. This indicates
* what stage the op is in, query, exec, done, duplicate, failed. */
enum op_state state;
/*! The node that owns the remote operation */
char *originator;
/*! The local client id that initiated the fencing request */
char *client_id;
/*! The client's call_id that initiated the fencing request */
int client_callid;
/*! The name of client that initiated the fencing request */
char *client_name;
/*! List of the received query results for all the nodes in the cpg group */
GListPtr query_results;
/*! The original request that initiated the remote stonith operation */
xmlNode *request;
/*! The current topology level being executed */
guint level;
/*! The current operation phase being executed */
enum st_remap_phase phase;
/*! Devices with automatic unfencing (always run if "on" requested, never if remapped) */
GListPtr automatic_list;
/*! List of all devices at the currently executing topology level */
GListPtr devices_list;
/*! Current entry in the topology device list */
GListPtr devices;
/*! List of duplicate operations attached to this operation. Once this operation
* completes, the duplicate operations will be closed out as well. */
GListPtr duplicates;
} remote_fencing_op_t;
/*!
* \internal
* \brief Broadcast the result of an operation to the peers.
* \param op, Operation whose result should be broadcast
* \param rc, Result of the operation
*/
void stonith_bcast_result_to_peers(remote_fencing_op_t * op, int rc);
enum st_callback_flags {
- st_callback_unknown = 0x0000,
- st_callback_notify_fence = 0x0001,
- st_callback_device_add = 0x0004,
- st_callback_device_del = 0x0010,
- st_callback_notify_history = 0x0020
+ st_callback_unknown = 0x0000,
+ st_callback_notify_fence = 0x0001,
+ st_callback_device_add = 0x0004,
+ st_callback_device_del = 0x0010,
+ st_callback_notify_history = 0x0020,
+ st_callback_notify_history_synced = 0x0040
};
/*
* Complex fencing requirements are specified via fencing topologies.
* A topology consists of levels; each level is a list of fencing devices.
* Topologies are stored in a hash table by node name. When a node needs to be
* fenced, if it has an entry in the topology table, the levels are tried
* sequentially, and the devices in each level are tried sequentially.
* Fencing is considered successful as soon as any level succeeds;
* a level is considered successful if all its devices succeed.
* Essentially, all devices at a given level are "and-ed" and the
* levels are "or-ed".
*
* This structure is used for the topology table entries.
* Topology levels start from 1, so levels[0] is unused and always NULL.
*/
typedef struct stonith_topology_s {
int kind;
/*! Node name regex or attribute name=value for which topology applies */
char *target;
char *target_value;
char *target_pattern;
char *target_attribute;
/*! Names of fencing devices at each topology level */
GListPtr levels[ST_LEVEL_MAX];
} stonith_topology_t;
void init_device_list(void);
void free_device_list(void);
void init_topology_list(void);
void free_topology_list(void);
void free_stonith_remote_op_list(void);
void init_stonith_remote_op_hash_table(GHashTable **table);
void free_metadata_cache(void);
long long get_stonith_flag(const char *name);
void stonith_command(crm_client_t * client, uint32_t id, uint32_t flags,
xmlNode * op_request, const char *remote_peer);
int stonith_device_register(xmlNode * msg, const char **desc, gboolean from_cib);
int stonith_device_remove(const char *id, gboolean from_cib);
char *stonith_level_key(xmlNode * msg, int mode);
int stonith_level_kind(xmlNode * msg);
int stonith_level_register(xmlNode * msg, char **desc);
int stonith_level_remove(xmlNode * msg, char **desc);
stonith_topology_t *find_topology_for_host(const char *host);
void do_local_reply(xmlNode * notify_src, const char *client_id, gboolean sync_reply,
gboolean from_peer);
xmlNode *stonith_construct_reply(xmlNode * request, const char *output, xmlNode * data,
int rc);
void
do_stonith_async_timeout_update(const char *client, const char *call_id, int timeout);
void do_stonith_notify(int options, const char *type, int result, xmlNode * data);
void do_stonith_notify_device(int options, const char *op, int rc, const char *desc);
void do_stonith_notify_level(int options, const char *op, int rc, const char *desc);
remote_fencing_op_t *initiate_remote_stonith_op(crm_client_t * client, xmlNode * request,
gboolean manual_ack);
int process_remote_stonith_exec(xmlNode * msg);
int process_remote_stonith_query(xmlNode * msg);
void *create_remote_stonith_op(const char *client, xmlNode * request, gboolean peer);
int stonith_fence_history(xmlNode *msg, xmlNode **output,
const char *remote_peer, int options);
void stonith_fence_history_trim(void);
bool fencing_peer_active(crm_node_t *peer);
int stonith_manual_ack(xmlNode * msg, remote_fencing_op_t * op);
gboolean string_in_list(GListPtr list, const char *item);
gboolean node_has_attr(const char *node, const char *name, const char *value);
extern char *stonith_our_uname;
extern gboolean stand_alone;
extern GHashTable *device_list;
extern GHashTable *topology;
extern long stonith_watchdog_timeout_ms;
extern GHashTable *stonith_remote_op_list;
diff --git a/include/crm/stonith-ng.h b/include/crm/stonith-ng.h
index b6407324df..418a03c85e 100644
--- a/include/crm/stonith-ng.h
+++ b/include/crm/stonith-ng.h
@@ -1,554 +1,555 @@
/*
* Copyright 2004-2019 the Pacemaker project contributors
*
* The version control history for this file may have further details.
*
* This source code is licensed under the GNU Lesser General Public License
* version 2.1 or later (LGPLv2.1+) WITHOUT ANY WARRANTY.
*/
#ifdef __cplusplus
extern "C" {
#endif
/**
* \file
* \brief Fencing aka. STONITH
* \ingroup fencing
*/
#ifndef STONITH_NG__H
# define STONITH_NG__H
# include <dlfcn.h>
# include <errno.h>
# include <stdbool.h> // bool
# include <stdint.h> // uint32_t
# include <time.h> // time_t
# define T_STONITH_NOTIFY_DISCONNECT "st_notify_disconnect"
# define T_STONITH_NOTIFY_FENCE "st_notify_fence"
# define T_STONITH_NOTIFY_HISTORY "st_notify_history"
+# define T_STONITH_NOTIFY_HISTORY_SYNCED "st_notify_history_synced"
/* *INDENT-OFF* */
enum stonith_state {
stonith_connected_command,
stonith_connected_query,
stonith_disconnected,
};
enum stonith_call_options {
st_opt_none = 0x00000000,
st_opt_verbose = 0x00000001,
st_opt_allow_suicide = 0x00000002,
st_opt_manual_ack = 0x00000008,
st_opt_discard_reply = 0x00000010,
/* st_opt_all_replies = 0x00000020, */
st_opt_topology = 0x00000040,
st_opt_scope_local = 0x00000100,
st_opt_cs_nodeid = 0x00000200,
st_opt_sync_call = 0x00001000,
/*! Allow the timeout period for a callback to be adjusted
* based on the time the server reports the operation will take. */
st_opt_timeout_updates = 0x00002000,
/*! Only report back if operation is a success in callback */
st_opt_report_only_success = 0x00004000,
/* used where ever apropriate - e.g. cleanup of history */
st_opt_cleanup = 0x000080000,
/* used where ever apropriate - e.g. send out a history query to all nodes */
st_opt_broadcast = 0x000100000,
};
/*! Order matters here, do not change values */
enum op_state
{
st_query,
st_exec,
st_done,
st_duplicate,
st_failed,
};
// Supported fence agent interface standards
enum stonith_namespace {
st_namespace_invalid,
st_namespace_any,
st_namespace_internal, // Implemented internally by Pacemaker
/* Neither of these projects are active any longer, but the fence agent
* interfaces they created are still in use and supported by Pacemaker.
*/
st_namespace_rhcs, // Red Hat Cluster Suite compatible
st_namespace_lha, // Linux-HA compatible
};
enum stonith_namespace stonith_text2namespace(const char *namespace_s);
const char *stonith_namespace2text(enum stonith_namespace st_namespace);
enum stonith_namespace stonith_get_namespace(const char *agent,
const char *namespace_s);
typedef struct stonith_key_value_s {
char *key;
char *value;
struct stonith_key_value_s *next;
} stonith_key_value_t;
typedef struct stonith_history_s {
char *target;
char *action;
char *origin;
char *delegate;
char *client;
int state;
time_t completed;
struct stonith_history_s *next;
} stonith_history_t;
typedef struct stonith_s stonith_t;
typedef struct stonith_event_s
{
char *id;
char *type;
char *message;
char *operation;
int result;
char *origin;
char *target;
char *action;
char *executioner;
char *device;
/*! The name of the client that initiated the action. */
char *client_origin;
} stonith_event_t;
typedef struct stonith_callback_data_s
{
int rc;
int call_id;
void *userdata;
} stonith_callback_data_t;
typedef struct stonith_api_operations_s
{
/*!
* \brief Destroy the stonith api structure.
*/
int (*free) (stonith_t *st);
/*!
* \brief Connect to the local stonith daemon.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*connect) (stonith_t *st, const char *name, int *stonith_fd);
/*!
* \brief Disconnect from the local stonith daemon.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*disconnect)(stonith_t *st);
/*!
* \brief Remove a registered stonith device with the local stonith daemon.
*
* \note Synchronous, guaranteed to occur in daemon before function returns.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*remove_device)(
stonith_t *st, int options, const char *name);
/*!
* \brief Register a stonith device with the local stonith daemon.
*
* \note Synchronous, guaranteed to occur in daemon before function returns.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*register_device)(
stonith_t *st, int options, const char *id,
const char *provider, const char *agent, stonith_key_value_t *params);
/*!
* \brief Remove a fencing level for a specific node.
*
* \note This feature is not available when stonith is in standalone mode.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*remove_level)(
stonith_t *st, int options, const char *node, int level);
/*!
* \brief Register a fencing level containing the fencing devices to be used
* at that level for a specific node.
*
* \note This feature is not available when stonith is in standalone mode.
*
* \retval 0, success
* \retval negative error code on failure
*/
int (*register_level)(
stonith_t *st, int options, const char *node, int level, stonith_key_value_t *device_list);
/*!
* \brief Get the metadata documentation for a resource.
*
* \note Value is returned in output. Output must be freed when set.
*
* \retval 0 success
* \retval negative error code on failure
*/
int (*metadata)(stonith_t *st, int options,
const char *device, const char *provider, char **output, int timeout);
/*!
* \brief Retrieve a list of installed stonith agents
*
* \note if provider is not provided, all known agents will be returned
* \note list must be freed using stonith_key_value_freeall()
* \note call_options parameter is not used, it is reserved for future use.
*
* \retval num items in list on success
* \retval negative error code on failure
*/
int (*list_agents)(stonith_t *stonith, int call_options, const char *provider,
stonith_key_value_t **devices, int timeout);
/*!
* \brief Retrieve string listing hosts and port assignments from a local stonith device.
*
* \retval 0 on success
* \retval negative error code on failure
*/
int (*list)(stonith_t *st, int options, const char *id, char **list_output, int timeout);
/*!
* \brief Check to see if a local stonith device is reachable
*
* \retval 0 on success
* \retval negative error code on failure
*/
int (*monitor)(stonith_t *st, int options, const char *id, int timeout);
/*!
* \brief Check to see if a local stonith device's port is reachable
*
* \retval 0 on success
* \retval negative error code on failure
*/
int (*status)(stonith_t *st, int options, const char *id, const char *port, int timeout);
/*!
* \brief Retrieve a list of registered stonith devices.
*
* \note If node is provided, only devices that can fence the node id
* will be returned.
*
* \retval num items in list on success
* \retval negative error code on failure
*/
int (*query)(stonith_t *st, int options, const char *node,
stonith_key_value_t **devices, int timeout);
/*!
* \brief Issue a fencing action against a node.
*
* \note Possible actions are, 'on', 'off', and 'reboot'.
*
* \param st, stonith connection
* \param options, call options
* \param node, The target node to fence
* \param action, The fencing action to take
* \param timeout, The default per device timeout to use with each device
* capable of fencing the target.
*
* \retval 0 success
* \retval negative error code on failure.
*/
int (*fence)(stonith_t *st, int options, const char *node, const char *action,
int timeout, int tolerance);
/*!
* \brief Manually confirm that a node is down.
*
* \retval 0 success
* \retval negative error code on failure.
*/
int (*confirm)(stonith_t *st, int options, const char *node);
/*!
* \brief Retrieve a list of fencing operations that have occurred for a specific node.
*
* \note History is not available in standalone mode.
*
* \retval 0 success
* \retval negative error code on failure.
*/
int (*history)(stonith_t *st, int options, const char *node, stonith_history_t **output, int timeout);
int (*register_notification)(
stonith_t *st, const char *event,
void (*notify)(stonith_t *st, stonith_event_t *e));
int (*remove_notification)(stonith_t *st, const char *event);
/*!
* \brief Register a callback to receive the result of an asynchronous call
*
* \param[in] call_id The call ID to register callback for
* \param[in] timeout Default time to wait until callback expires
* \param[in] options Bitmask of \c stonith_call_options (respects
* \c st_opt_timeout_updates and
* \c st_opt_report_only_success)
* \param[in] userdata Pointer that will be given to callback
* \param[in] callback_name Unique name to identify callback
* \param[in] callback The callback function to register
*
* \return \c TRUE on success, \c FALSE if call_id is negative, -errno otherwise
*
* \todo This function should return \c pcmk_ok on success, and \c call_id
* when negative, but that would break backward compatibility.
*/
int (*register_callback)(stonith_t *st,
int call_id,
int timeout,
int options,
void *userdata,
const char *callback_name,
void (*callback)(stonith_t *st, stonith_callback_data_t *data));
/*!
* \brief Remove a registered callback for a given call id.
*/
int (*remove_callback)(stonith_t *st, int call_id, bool all_callbacks);
/*!
* \brief Remove fencing level for specific node, node regex or attribute
*
* \param[in] st Fencer connection to use
* \param[in] options Bitmask of stonith_call_options to pass to the fencer
* \param[in] node If not NULL, target level by this node name
* \param[in] pattern If not NULL, target by node name using this regex
* \param[in] attr If not NULL, target by this node attribute
* \param[in] value If not NULL, target by this node attribute value
* \param[in] level Index number of level to remove
*
* \return 0 on success, negative error code otherwise
*
* \note This feature is not available when stonith is in standalone mode.
* The caller should set only one of node, pattern or attr/value.
*/
int (*remove_level_full)(stonith_t *st, int options,
const char *node, const char *pattern,
const char *attr, const char *value, int level);
/*!
* \brief Register fencing level for specific node, node regex or attribute
*
* \param[in] st Fencer connection to use
* \param[in] options Bitmask of stonith_call_options to pass to fencer
* \param[in] node If not NULL, target level by this node name
* \param[in] pattern If not NULL, target by node name using this regex
* \param[in] attr If not NULL, target by this node attribute
* \param[in] value If not NULL, target by this node attribute value
* \param[in] level Index number of level to add
* \param[in] device_list Devices to use in level
*
* \return 0 on success, negative error code otherwise
*
* \note This feature is not available when stonith is in standalone mode.
* The caller should set only one of node, pattern or attr/value.
*/
int (*register_level_full)(stonith_t *st, int options,
const char *node, const char *pattern,
const char *attr, const char *value,
int level, stonith_key_value_t *device_list);
/*!
* \brief Validate an arbitrary stonith device configuration
*
* \param[in] st Stonithd connection to use
* \param[in] call_options Bitmask of stonith_call_options to use with fencer
* \param[in] rsc_id ID used to replace CIB secrets in params
* \param[in] namespace_s Namespace of fence agent to validate (optional)
* \param[in] agent Fence agent to validate
* \param[in] params Configuration parameters to pass to fence agent
* \param[in] timeout Fail if no response within this many seconds
* \param[out] output If non-NULL, where to store any agent output
* \param[out] error_output If non-NULL, where to store agent error output
*
* \return pcmk_ok if validation succeeds, -errno otherwise
*
* \note If pcmk_ok is returned, the caller is responsible for freeing
* the output (if requested).
*/
int (*validate)(stonith_t *st, int call_options, const char *rsc_id,
const char *namespace_s, const char *agent,
stonith_key_value_t *params, int timeout, char **output,
char **error_output);
} stonith_api_operations_t;
struct stonith_s
{
enum stonith_state state;
int call_id;
int call_timeout;
void *st_private;
stonith_api_operations_t *cmds;
};
/* *INDENT-ON* */
/* Core functions */
stonith_t *stonith_api_new(void);
void stonith_api_delete(stonith_t * st);
void stonith_dump_pending_callbacks(stonith_t * st);
// deprecated (use stonith_get_namespace() instead)
const char *get_stonith_provider(const char *agent, const char *provider);
bool stonith_dispatch(stonith_t * st);
stonith_key_value_t *stonith_key_value_add(stonith_key_value_t * kvp, const char *key,
const char *value);
void stonith_key_value_freeall(stonith_key_value_t * kvp, int keys, int values);
void stonith_history_free(stonith_history_t *history);
// Convenience functions
int stonith_api_connect_retry(stonith_t *st, const char *name,
int max_attempts);
/* Basic helpers that allows nodes to be fenced and the history to be
* queried without mainloop or the caller understanding the full API
*
* At least one of nodeid and uname are required
*/
int stonith_api_kick(uint32_t nodeid, const char *uname, int timeout, bool off);
time_t stonith_api_time(uint32_t nodeid, const char *uname, bool in_progress);
/*
* Helpers for using the above functions without install-time dependencies
*
* Usage:
* #include <crm/stonith-ng.h>
*
* To turn a node off by corosync nodeid:
* stonith_api_kick_helper(nodeid, 120, 1);
*
* To check the last fence date/time (also by nodeid):
* last = stonith_api_time_helper(nodeid, 0);
*
* To check if fencing is in progress:
* if(stonith_api_time_helper(nodeid, 1) > 0) { ... }
*
* eg.
#include <stdio.h>
#include <time.h>
#include <crm/stonith-ng.h>
int
main(int argc, char ** argv)
{
int rc = 0;
int nodeid = 102;
rc = stonith_api_time_helper(nodeid, 0);
printf("%d last fenced at %s\n", nodeid, ctime(rc));
rc = stonith_api_kick_helper(nodeid, 120, 1);
printf("%d fence result: %d\n", nodeid, rc);
rc = stonith_api_time_helper(nodeid, 0);
printf("%d last fenced at %s\n", nodeid, ctime(rc));
return 0;
}
*/
# define STONITH_LIBRARY "libstonithd.so.26"
typedef int (*st_api_kick_fn) (int nodeid, const char *uname, int timeout, bool off);
typedef time_t (*st_api_time_fn) (int nodeid, const char *uname, bool in_progress);
static inline int
stonith_api_kick_helper(uint32_t nodeid, int timeout, bool off)
{
static void *st_library = NULL;
static st_api_kick_fn st_kick_fn;
if (st_library == NULL) {
st_library = dlopen(STONITH_LIBRARY, RTLD_LAZY);
}
if (st_library && st_kick_fn == NULL) {
st_kick_fn = (st_api_kick_fn) dlsym(st_library, "stonith_api_kick");
}
if (st_kick_fn == NULL) {
#ifdef ELIBACC
return -ELIBACC;
#else
return -ENOSYS;
#endif
}
return (*st_kick_fn) (nodeid, NULL, timeout, off);
}
static inline time_t
stonith_api_time_helper(uint32_t nodeid, bool in_progress)
{
static void *st_library = NULL;
static st_api_time_fn st_time_fn;
if (st_library == NULL) {
st_library = dlopen(STONITH_LIBRARY, RTLD_LAZY);
}
if (st_library && st_time_fn == NULL) {
st_time_fn = (st_api_time_fn) dlsym(st_library, "stonith_api_time");
}
if (st_time_fn == NULL) {
return 0;
}
return (*st_time_fn) (nodeid, NULL, in_progress);
}
/**
* Does the given agent describe a stonith resource that can exist?
*
* \param[in] agent What is the name of the agent?
* \param[in] timeout Timeout to use when querying. If 0 is given,
* use a default of 120.
*
* \return A boolean
*/
bool stonith_agent_exists(const char *agent, int timeout);
/*!
* \brief Turn stonith action into a more readable string.
*
* \param action Stonith action
*/
const char *stonith_action_str(const char *action);
#ifdef __cplusplus
}
#endif
#endif

File Metadata

Mime Type
text/x-diff
Expires
Sat, Nov 23, 4:47 PM (17 h, 2 m)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
1018939
Default Alt Text
(96 KB)

Event Timeline