Page MenuHomeClusterLabs Projects
Files F1841690

No OneTemporary
Actions

View Options

This file is larger than 256 KB, so syntax highlighting was skipped.
diff --git a/daemons/controld/controld_te_utils.c b/daemons/controld/controld_te_utils.c
index 847c5d39c9..7cb94415ac 100644
--- a/daemons/controld/controld_te_utils.c
+++ b/daemons/controld/controld_te_utils.c
@@ -1,293 +1,292 @@
/*
* 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 General Public License version 2
* or later (GPLv2+) WITHOUT ANY WARRANTY.
*/
#include <crm_internal.h>
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <pacemaker-controld.h>
gboolean
stop_te_timer(crm_action_timer_t * timer)
{
if (timer == NULL) {
return FALSE;
}
if (timer->source_id != 0) {
crm_trace("Stopping action timer");
g_source_remove(timer->source_id);
timer->source_id = 0;
} else {
crm_trace("Action timer was already stopped");
return FALSE;
}
return TRUE;
}
gboolean
te_graph_trigger(gpointer user_data)
{
enum transition_status graph_rc = -1;
if (transition_graph == NULL) {
crm_debug("Nothing to do");
return TRUE;
}
crm_trace("Invoking graph %d in state %s", transition_graph->id, fsa_state2string(fsa_state));
switch (fsa_state) {
case S_STARTING:
case S_PENDING:
case S_NOT_DC:
case S_HALT:
case S_ILLEGAL:
case S_STOPPING:
case S_TERMINATE:
return TRUE;
- break;
default:
break;
}
if (transition_graph->complete == FALSE) {
int limit = transition_graph->batch_limit;
transition_graph->batch_limit = throttle_get_total_job_limit(limit);
graph_rc = run_graph(transition_graph);
transition_graph->batch_limit = limit; /* Restore the configured value */
/* significant overhead... */
/* print_graph(LOG_TRACE, transition_graph); */
if (graph_rc == transition_active) {
crm_trace("Transition not yet complete");
return TRUE;
} else if (graph_rc == transition_pending) {
crm_trace("Transition not yet complete - no actions fired");
return TRUE;
}
if (graph_rc != transition_complete) {
crm_warn("Transition failed: %s", transition_status(graph_rc));
print_graph(LOG_NOTICE, transition_graph);
}
}
crm_debug("Transition %d is now complete", transition_graph->id);
transition_graph->complete = TRUE;
notify_crmd(transition_graph);
return TRUE;
}
void
trigger_graph_processing(const char *fn, int line)
{
crm_trace("%s:%d - Triggered graph processing", fn, line);
mainloop_set_trigger(transition_trigger);
}
static struct abort_timer_s {
bool aborted;
guint id;
int priority;
enum transition_action action;
const char *text;
} abort_timer = { 0, };
static gboolean
abort_timer_popped(gpointer data)
{
if (AM_I_DC && (abort_timer.aborted == FALSE)) {
abort_transition(abort_timer.priority, abort_timer.action,
abort_timer.text, NULL);
}
abort_timer.id = 0;
return FALSE; // do not immediately reschedule timer
}
/*!
* \internal
* \brief Abort transition after delay, if not already aborted in that time
*
* \param[in] abort_text Must be literal string
*/
void
abort_after_delay(int abort_priority, enum transition_action abort_action,
const char *abort_text, guint delay_ms)
{
if (abort_timer.id) {
// Timer already in progress, stop and reschedule
g_source_remove(abort_timer.id);
}
abort_timer.aborted = FALSE;
abort_timer.priority = abort_priority;
abort_timer.action = abort_action;
abort_timer.text = abort_text;
abort_timer.id = g_timeout_add(delay_ms, abort_timer_popped, NULL);
}
void
abort_transition_graph(int abort_priority, enum transition_action abort_action,
const char *abort_text, xmlNode * reason, const char *fn, int line)
{
int add[] = { 0, 0, 0 };
int del[] = { 0, 0, 0 };
int level = LOG_INFO;
xmlNode *diff = NULL;
xmlNode *change = NULL;
CRM_CHECK(transition_graph != NULL, return);
switch (fsa_state) {
case S_STARTING:
case S_PENDING:
case S_NOT_DC:
case S_HALT:
case S_ILLEGAL:
case S_STOPPING:
case S_TERMINATE:
crm_info("Abort %s suppressed: state=%s (complete=%d)",
abort_text, fsa_state2string(fsa_state), transition_graph->complete);
return;
default:
break;
}
abort_timer.aborted = TRUE;
controld_expect_sched_reply(NULL);
if (transition_graph->complete == FALSE) {
if(update_abort_priority(transition_graph, abort_priority, abort_action, abort_text)) {
level = LOG_NOTICE;
}
}
if(reason) {
xmlNode *search = NULL;
for(search = reason; search; search = search->parent) {
if (safe_str_eq(XML_TAG_DIFF, TYPE(search))) {
diff = search;
break;
}
}
if(diff) {
xml_patch_versions(diff, add, del);
for(search = reason; search; search = search->parent) {
if (safe_str_eq(XML_DIFF_CHANGE, TYPE(search))) {
change = search;
break;
}
}
}
}
if(reason == NULL) {
do_crm_log(level, "Transition %d aborted: %s "CRM_XS" source=%s:%d complete=%s",
transition_graph->id, abort_text, fn, line,
(transition_graph->complete? "true" : "false"));
} else if(change == NULL) {
char *local_path = xml_get_path(reason);
do_crm_log(level, "Transition %d aborted by %s.%s: %s "
CRM_XS " cib=%d.%d.%d source=%s:%d path=%s complete=%s",
transition_graph->id, TYPE(reason), ID(reason), abort_text,
add[0], add[1], add[2], fn, line, local_path,
(transition_graph->complete? "true" : "false"));
free(local_path);
} else {
const char *kind = NULL;
const char *op = crm_element_value(change, XML_DIFF_OP);
const char *path = crm_element_value(change, XML_DIFF_PATH);
if(change == reason) {
if(strcmp(op, "create") == 0) {
reason = reason->children;
} else if(strcmp(op, "modify") == 0) {
reason = first_named_child(reason, XML_DIFF_RESULT);
if(reason) {
reason = reason->children;
}
}
}
kind = TYPE(reason);
if(strcmp(op, "delete") == 0) {
const char *shortpath = strrchr(path, '/');
do_crm_log(level, "Transition %d aborted by deletion of %s: %s "
CRM_XS " cib=%d.%d.%d source=%s:%d path=%s complete=%s",
transition_graph->id,
(shortpath? (shortpath + 1) : path), abort_text,
add[0], add[1], add[2], fn, line, path,
(transition_graph->complete? "true" : "false"));
} else if (safe_str_eq(XML_CIB_TAG_NVPAIR, kind)) {
do_crm_log(level, "Transition %d aborted by %s doing %s %s=%s: %s "
CRM_XS " cib=%d.%d.%d source=%s:%d path=%s complete=%s",
transition_graph->id,
crm_element_value(reason, XML_ATTR_ID), op,
crm_element_value(reason, XML_NVPAIR_ATTR_NAME),
crm_element_value(reason, XML_NVPAIR_ATTR_VALUE),
abort_text, add[0], add[1], add[2], fn, line, path,
(transition_graph->complete? "true" : "false"));
} else if (safe_str_eq(XML_LRM_TAG_RSC_OP, kind)) {
const char *magic = crm_element_value(reason, XML_ATTR_TRANSITION_MAGIC);
do_crm_log(level, "Transition %d aborted by operation %s '%s' on %s: %s "
CRM_XS " magic=%s cib=%d.%d.%d source=%s:%d complete=%s",
transition_graph->id,
crm_element_value(reason, XML_LRM_ATTR_TASK_KEY), op,
crm_element_value(reason, XML_LRM_ATTR_TARGET), abort_text,
magic, add[0], add[1], add[2], fn, line,
(transition_graph->complete? "true" : "false"));
} else if (pcmk__str_any_of(kind, XML_CIB_TAG_STATE, XML_CIB_TAG_NODE, NULL)) {
const char *uname = crm_peer_uname(ID(reason));
do_crm_log(level, "Transition %d aborted by %s '%s' on %s: %s "
CRM_XS " cib=%d.%d.%d source=%s:%d complete=%s",
transition_graph->id,
kind, op, (uname? uname : ID(reason)), abort_text,
add[0], add[1], add[2], fn, line,
(transition_graph->complete? "true" : "false"));
} else {
const char *id = ID(reason);
do_crm_log(level, "Transition %d aborted by %s.%s '%s': %s "
CRM_XS " cib=%d.%d.%d source=%s:%d path=%s complete=%s",
transition_graph->id,
TYPE(reason), (id? id : ""), (op? op : "change"),
abort_text, add[0], add[1], add[2], fn, line, path,
(transition_graph->complete? "true" : "false"));
}
}
if (transition_graph->complete) {
if (transition_timer->period_ms > 0) {
controld_stop_timer(transition_timer);
controld_start_timer(transition_timer);
} else {
register_fsa_input(C_FSA_INTERNAL, I_PE_CALC, NULL);
}
return;
}
mainloop_set_trigger(transition_trigger);
}
diff --git a/lib/cluster/corosync.c b/lib/cluster/corosync.c
index 648c0d534f..f7b81ba567 100644
--- a/lib/cluster/corosync.c
+++ b/lib/cluster/corosync.c
@@ -1,767 +1,766 @@
/*
* 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.
*/
#include <crm_internal.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <arpa/inet.h>
#include <netdb.h>
#include <inttypes.h> /* U64T ~ PRIu64 */
#include <bzlib.h>
#include <crm/common/ipc.h>
#include <crm/cluster/internal.h>
#include <crm/common/mainloop.h>
#include <sys/utsname.h>
#include <qb/qbipcc.h>
#include <qb/qbutil.h>
#include <corosync/corodefs.h>
#include <corosync/corotypes.h>
#include <corosync/hdb.h>
#include <corosync/cfg.h>
#include <corosync/cmap.h>
#include <corosync/quorum.h>
#include <crm/msg_xml.h>
#include <crm/common/ipc_internal.h> /* PCMK__SPECIAL_PID* */
quorum_handle_t pcmk_quorum_handle = 0;
gboolean(*quorum_app_callback) (unsigned long long seq, gboolean quorate) = NULL;
char *
get_corosync_uuid(crm_node_t *node)
{
if (node && is_corosync_cluster()) {
if (node->id > 0) {
return crm_strdup_printf("%u", node->id);
} else {
crm_info("Node %s is not yet known by corosync", node->uname);
}
}
return NULL;
}
/*
* CFG functionality stolen from node_name() in corosync-quorumtool.c
* This resolves the first address assigned to a node and returns the name or IP address.
*/
char *
corosync_node_name(uint64_t /*cmap_handle_t */ cmap_handle, uint32_t nodeid)
{
int lpc = 0;
cs_error_t rc = CS_OK;
int retries = 0;
char *name = NULL;
cmap_handle_t local_handle = 0;
int fd = -1;
uid_t found_uid = 0;
gid_t found_gid = 0;
pid_t found_pid = 0;
int rv;
if (nodeid == 0) {
nodeid = get_local_nodeid(0);
}
if (cmap_handle == 0 && local_handle == 0) {
retries = 0;
crm_trace("Initializing CMAP connection");
do {
rc = cmap_initialize(&local_handle);
if (rc != CS_OK) {
retries++;
crm_debug("API connection setup failed: %s. Retrying in %ds", cs_strerror(rc),
retries);
sleep(retries);
}
} while (retries < 5 && rc != CS_OK);
if (rc != CS_OK) {
crm_warn("Could not connect to Cluster Configuration Database API, error %s",
cs_strerror(rc));
local_handle = 0;
}
}
if (cmap_handle == 0) {
cmap_handle = local_handle;
rc = cmap_fd_get(cmap_handle, &fd);
if (rc != CS_OK) {
crm_err("Could not obtain the CMAP API connection: %s (%d)",
cs_strerror(rc), rc);
goto bail;
}
/* CMAP provider run as root (in given user namespace, anyway)? */
if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid,
&found_uid, &found_gid))) {
crm_err("CMAP provider is not authentic:"
" process %lld (uid: %lld, gid: %lld)",
(long long) PCMK__SPECIAL_PID_AS_0(found_pid),
(long long) found_uid, (long long) found_gid);
goto bail;
} else if (rv < 0) {
crm_err("Could not verify authenticity of CMAP provider: %s (%d)",
strerror(-rv), -rv);
goto bail;
}
}
while (name == NULL && cmap_handle != 0) {
uint32_t id = 0;
char *key = NULL;
key = crm_strdup_printf("nodelist.node.%d.nodeid", lpc);
rc = cmap_get_uint32(cmap_handle, key, &id);
crm_trace("Checking %u vs %u from %s", nodeid, id, key);
free(key);
if (rc != CS_OK) {
break;
}
if (nodeid == id) {
crm_trace("Searching for node name for %u in nodelist.node.%d %s", nodeid, lpc, name);
if (name == NULL) {
key = crm_strdup_printf("nodelist.node.%d.name", lpc);
cmap_get_string(cmap_handle, key, &name);
crm_trace("%s = %s", key, name);
free(key);
}
if (name == NULL) {
key = crm_strdup_printf("nodelist.node.%d.ring0_addr", lpc);
cmap_get_string(cmap_handle, key, &name);
crm_trace("%s = %s", key, name);
if (node_name_is_valid(key, name) == FALSE) {
free(name);
name = NULL;
}
free(key);
}
break;
}
lpc++;
}
bail:
if(local_handle) {
cmap_finalize(local_handle);
}
if (name == NULL) {
crm_info("Unable to get node name for nodeid %u", nodeid);
}
return name;
}
void
terminate_cs_connection(crm_cluster_t *cluster)
{
cluster_disconnect_cpg(cluster);
if (pcmk_quorum_handle) {
quorum_finalize(pcmk_quorum_handle);
pcmk_quorum_handle = 0;
}
crm_notice("Disconnected from Corosync");
}
static int
pcmk_quorum_dispatch(gpointer user_data)
{
int rc = 0;
rc = quorum_dispatch(pcmk_quorum_handle, CS_DISPATCH_ALL);
if (rc < 0) {
crm_err("Connection to the Quorum API failed: %d", rc);
pcmk_quorum_handle = 0;
return -1;
}
return 0;
}
static void
pcmk_quorum_notification(quorum_handle_t handle,
uint32_t quorate,
uint64_t ring_id, uint32_t view_list_entries, uint32_t * view_list)
{
int i;
GHashTableIter iter;
crm_node_t *node = NULL;
static gboolean init_phase = TRUE;
if (quorate != crm_have_quorum) {
if (quorate) {
crm_notice("Quorum acquired " CRM_XS " membership=%" U64T " members=%lu",
ring_id, (long unsigned int)view_list_entries);
} else {
crm_warn("Quorum lost " CRM_XS " membership=%" U64T " members=%lu",
ring_id, (long unsigned int)view_list_entries);
}
crm_have_quorum = quorate;
} else {
crm_info("Quorum %s " CRM_XS " membership=%" U64T " members=%lu",
(quorate? "retained" : "still lost"), ring_id,
(long unsigned int)view_list_entries);
}
if (view_list_entries == 0 && init_phase) {
crm_info("Corosync membership is still forming, ignoring");
return;
}
init_phase = FALSE;
/* Reset last_seen for all cached nodes so we can tell which ones aren't
* in the view list */
g_hash_table_iter_init(&iter, crm_peer_cache);
while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) {
node->last_seen = 0;
}
/* Update the peer cache for each node in view list */
for (i = 0; i < view_list_entries; i++) {
uint32_t id = view_list[i];
crm_debug("Member[%d] %u ", i, id);
/* Get this node's peer cache entry (adding one if not already there) */
node = crm_get_peer(id, NULL);
if (node->uname == NULL) {
char *name = corosync_node_name(0, id);
crm_info("Obtaining name for new node %u", id);
node = crm_get_peer(id, name);
free(name);
}
/* Update the node state (including updating last_seen to ring_id) */
crm_update_peer_state(__FUNCTION__, node, CRM_NODE_MEMBER, ring_id);
}
/* Remove any peer cache entries we didn't update */
crm_reap_unseen_nodes(ring_id);
if (quorum_app_callback) {
quorum_app_callback(ring_id, quorate);
}
}
quorum_callbacks_t quorum_callbacks = {
.quorum_notify_fn = pcmk_quorum_notification,
};
gboolean
cluster_connect_quorum(gboolean(*dispatch) (unsigned long long, gboolean),
void (*destroy) (gpointer))
{
cs_error_t rc;
int fd = 0;
int quorate = 0;
uint32_t quorum_type = 0;
struct mainloop_fd_callbacks quorum_fd_callbacks;
uid_t found_uid = 0;
gid_t found_gid = 0;
pid_t found_pid = 0;
int rv;
quorum_fd_callbacks.dispatch = pcmk_quorum_dispatch;
quorum_fd_callbacks.destroy = destroy;
crm_debug("Configuring Pacemaker to obtain quorum from Corosync");
rc = quorum_initialize(&pcmk_quorum_handle, &quorum_callbacks, &quorum_type);
if (rc != CS_OK) {
crm_err("Could not connect to the Quorum API: %s (%d)",
cs_strerror(rc), rc);
goto bail;
} else if (quorum_type != QUORUM_SET) {
crm_err("Corosync quorum is not configured");
goto bail;
}
rc = quorum_fd_get(pcmk_quorum_handle, &fd);
if (rc != CS_OK) {
crm_err("Could not obtain the Quorum API connection: %s (%d)",
strerror(rc), rc);
goto bail;
}
/* Quorum provider run as root (in given user namespace, anyway)? */
if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid,
&found_uid, &found_gid))) {
crm_err("Quorum provider is not authentic:"
" process %lld (uid: %lld, gid: %lld)",
(long long) PCMK__SPECIAL_PID_AS_0(found_pid),
(long long) found_uid, (long long) found_gid);
rc = CS_ERR_ACCESS;
goto bail;
} else if (rv < 0) {
crm_err("Could not verify authenticity of Quorum provider: %s (%d)",
strerror(-rv), -rv);
rc = CS_ERR_ACCESS;
goto bail;
}
rc = quorum_getquorate(pcmk_quorum_handle, &quorate);
if (rc != CS_OK) {
crm_err("Could not obtain the current Quorum API state: %d", rc);
goto bail;
}
if (quorate) {
crm_notice("Quorum acquired");
} else {
crm_warn("Quorum lost");
}
quorum_app_callback = dispatch;
crm_have_quorum = quorate;
rc = quorum_trackstart(pcmk_quorum_handle, CS_TRACK_CHANGES | CS_TRACK_CURRENT);
if (rc != CS_OK) {
crm_err("Could not setup Quorum API notifications: %d", rc);
goto bail;
}
mainloop_add_fd("quorum", G_PRIORITY_HIGH, fd, dispatch, &quorum_fd_callbacks);
corosync_initialize_nodelist(NULL, FALSE, NULL);
bail:
if (rc != CS_OK) {
quorum_finalize(pcmk_quorum_handle);
return FALSE;
}
return TRUE;
}
gboolean
init_cs_connection(crm_cluster_t * cluster)
{
int retries = 0;
while (retries < 5) {
int rc = init_cs_connection_once(cluster);
retries++;
switch (rc) {
case CS_OK:
return TRUE;
- break;
case CS_ERR_TRY_AGAIN:
case CS_ERR_QUEUE_FULL:
sleep(retries);
break;
default:
return FALSE;
}
}
crm_err("Could not connect to corosync after %d retries", retries);
return FALSE;
}
gboolean
init_cs_connection_once(crm_cluster_t * cluster)
{
crm_node_t *peer = NULL;
enum cluster_type_e stack = get_cluster_type();
crm_peer_init();
/* Here we just initialize comms */
if (stack != pcmk_cluster_corosync) {
crm_err("Invalid cluster type: %s (%d)", name_for_cluster_type(stack), stack);
return FALSE;
}
if (cluster_connect_cpg(cluster) == FALSE) {
return FALSE;
}
crm_info("Connection to '%s': established", name_for_cluster_type(stack));
cluster->nodeid = get_local_nodeid(0);
if(cluster->nodeid == 0) {
crm_err("Could not establish local nodeid");
return FALSE;
}
cluster->uname = get_node_name(0);
if(cluster->uname == NULL) {
crm_err("Could not establish local node name");
return FALSE;
}
/* Ensure the local node always exists */
peer = crm_get_peer(cluster->nodeid, cluster->uname);
cluster->uuid = get_corosync_uuid(peer);
return TRUE;
}
gboolean
check_message_sanity(const AIS_Message * msg, const char *data)
{
gboolean sane = TRUE;
int dest = msg->host.type;
int tmp_size = msg->header.size - sizeof(AIS_Message);
if (sane && msg->header.size == 0) {
crm_warn("Message with no size");
sane = FALSE;
}
if (sane && msg->header.error != CS_OK) {
crm_warn("Message header contains an error: %d", msg->header.error);
sane = FALSE;
}
if (sane && ais_data_len(msg) != tmp_size) {
crm_warn("Message payload size is incorrect: expected %d, got %d", ais_data_len(msg),
tmp_size);
sane = TRUE;
}
if (sane && ais_data_len(msg) == 0) {
crm_warn("Message with no payload");
sane = FALSE;
}
if (sane && data && msg->is_compressed == FALSE) {
int str_size = strlen(data) + 1;
if (ais_data_len(msg) != str_size) {
int lpc = 0;
crm_warn("Message payload is corrupted: expected %d bytes, got %d",
ais_data_len(msg), str_size);
sane = FALSE;
for (lpc = (str_size - 10); lpc < msg->size; lpc++) {
if (lpc < 0) {
lpc = 0;
}
crm_debug("bad_data[%d]: %d / '%c'", lpc, data[lpc], data[lpc]);
}
}
}
if (sane == FALSE) {
crm_err("Invalid message %d: (dest=%s:%s, from=%s:%s.%u, compressed=%d, size=%d, total=%d)",
msg->id, ais_dest(&(msg->host)), msg_type2text(dest),
ais_dest(&(msg->sender)), msg_type2text(msg->sender.type),
msg->sender.pid, msg->is_compressed, ais_data_len(msg), msg->header.size);
} else {
crm_trace
("Verified message %d: (dest=%s:%s, from=%s:%s.%u, compressed=%d, size=%d, total=%d)",
msg->id, ais_dest(&(msg->host)), msg_type2text(dest), ais_dest(&(msg->sender)),
msg_type2text(msg->sender.type), msg->sender.pid, msg->is_compressed,
ais_data_len(msg), msg->header.size);
}
return sane;
}
enum cluster_type_e
find_corosync_variant(void)
{
int rc = CS_OK;
cmap_handle_t handle;
rc = cmap_initialize(&handle);
switch(rc) {
case CS_OK:
break;
case CS_ERR_SECURITY:
crm_debug("Failed to initialize the cmap API: Permission denied (%d)", rc);
/* It's there, we just can't talk to it.
* Good enough for us to identify as 'corosync'
*/
return pcmk_cluster_corosync;
default:
crm_info("Failed to initialize the cmap API: %s (%d)",
ais_error2text(rc), rc);
return pcmk_cluster_unknown;
}
cmap_finalize(handle);
return pcmk_cluster_corosync;
}
gboolean
crm_is_corosync_peer_active(const crm_node_t * node)
{
if (node == NULL) {
crm_trace("NULL");
return FALSE;
} else if (safe_str_neq(node->state, CRM_NODE_MEMBER)) {
crm_trace("%s: state=%s", node->uname, node->state);
return FALSE;
} else if ((node->processes & crm_proc_cpg) == 0) {
crm_trace("%s: processes=%.16x", node->uname, node->processes);
return FALSE;
}
return TRUE;
}
gboolean
corosync_initialize_nodelist(void *cluster, gboolean force_member, xmlNode * xml_parent)
{
int lpc = 0;
cs_error_t rc = CS_OK;
int retries = 0;
gboolean any = FALSE;
cmap_handle_t cmap_handle;
int fd = -1;
uid_t found_uid = 0;
gid_t found_gid = 0;
pid_t found_pid = 0;
int rv;
do {
rc = cmap_initialize(&cmap_handle);
if (rc != CS_OK) {
retries++;
crm_debug("API connection setup failed: %s. Retrying in %ds", cs_strerror(rc),
retries);
sleep(retries);
}
} while (retries < 5 && rc != CS_OK);
if (rc != CS_OK) {
crm_warn("Could not connect to Cluster Configuration Database API, error %d", rc);
return FALSE;
}
rc = cmap_fd_get(cmap_handle, &fd);
if (rc != CS_OK) {
crm_err("Could not obtain the CMAP API connection: %s (%d)",
cs_strerror(rc), rc);
goto bail;
}
/* CMAP provider run as root (in given user namespace, anyway)? */
if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid,
&found_uid, &found_gid))) {
crm_err("CMAP provider is not authentic:"
" process %lld (uid: %lld, gid: %lld)",
(long long) PCMK__SPECIAL_PID_AS_0(found_pid),
(long long) found_uid, (long long) found_gid);
goto bail;
} else if (rv < 0) {
crm_err("Could not verify authenticity of CMAP provider: %s (%d)",
strerror(-rv), -rv);
goto bail;
}
crm_peer_init();
crm_trace("Initializing corosync nodelist");
for (lpc = 0; TRUE; lpc++) {
uint32_t nodeid = 0;
char *name = NULL;
char *key = NULL;
key = crm_strdup_printf("nodelist.node.%d.nodeid", lpc);
rc = cmap_get_uint32(cmap_handle, key, &nodeid);
free(key);
if (rc != CS_OK) {
break;
}
name = corosync_node_name(cmap_handle, nodeid);
if (name != NULL) {
GHashTableIter iter;
crm_node_t *node = NULL;
g_hash_table_iter_init(&iter, crm_peer_cache);
while (g_hash_table_iter_next(&iter, NULL, (gpointer *) &node)) {
if(node && node->uname && strcasecmp(node->uname, name) == 0) {
if (node->id && node->id != nodeid) {
crm_crit("Nodes %u and %u share the same name '%s': shutting down", node->id,
nodeid, name);
crm_exit(CRM_EX_FATAL);
}
}
}
}
if (nodeid > 0 || name != NULL) {
crm_trace("Initializing node[%d] %u = %s", lpc, nodeid, name);
crm_get_peer(nodeid, name);
}
if (nodeid > 0 && name != NULL) {
any = TRUE;
if (xml_parent) {
xmlNode *node = create_xml_node(xml_parent, XML_CIB_TAG_NODE);
crm_xml_set_id(node, "%u", nodeid);
crm_xml_add(node, XML_ATTR_UNAME, name);
if (force_member) {
crm_xml_add(node, XML_ATTR_TYPE, CRM_NODE_MEMBER);
}
}
}
free(name);
}
bail:
cmap_finalize(cmap_handle);
return any;
}
char *
corosync_cluster_name(void)
{
cmap_handle_t handle;
char *cluster_name = NULL;
cs_error_t rc = CS_OK;
int fd = -1;
uid_t found_uid = 0;
gid_t found_gid = 0;
pid_t found_pid = 0;
int rv;
rc = cmap_initialize(&handle);
if (rc != CS_OK) {
crm_info("Failed to initialize the cmap API: %s (%d)",
cs_strerror(rc), rc);
return NULL;
}
rc = cmap_fd_get(handle, &fd);
if (rc != CS_OK) {
crm_err("Could not obtain the CMAP API connection: %s (%d)",
cs_strerror(rc), rc);
goto bail;
}
/* CMAP provider run as root (in given user namespace, anyway)? */
if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid,
&found_uid, &found_gid))) {
crm_err("CMAP provider is not authentic:"
" process %lld (uid: %lld, gid: %lld)",
(long long) PCMK__SPECIAL_PID_AS_0(found_pid),
(long long) found_uid, (long long) found_gid);
goto bail;
} else if (rv < 0) {
crm_err("Could not verify authenticity of CMAP provider: %s (%d)",
strerror(-rv), -rv);
goto bail;
}
rc = cmap_get_string(handle, "totem.cluster_name", &cluster_name);
if (rc != CS_OK) {
crm_info("Cannot get totem.cluster_name: %s (%d)", cs_strerror(rc), rc);
} else {
crm_debug("cmap totem.cluster_name = '%s'", cluster_name);
}
bail:
cmap_finalize(handle);
return cluster_name;
}
int
corosync_cmap_has_config(const char *prefix)
{
cs_error_t rc = CS_OK;
int retries = 0;
static int found = -1;
cmap_handle_t cmap_handle;
cmap_iter_handle_t iter_handle;
char key_name[CMAP_KEYNAME_MAXLEN + 1];
int fd = -1;
uid_t found_uid = 0;
gid_t found_gid = 0;
pid_t found_pid = 0;
int rv;
if(found != -1) {
return found;
}
do {
rc = cmap_initialize(&cmap_handle);
if (rc != CS_OK) {
retries++;
crm_debug("API connection setup failed: %s. Retrying in %ds", cs_strerror(rc),
retries);
sleep(retries);
}
} while (retries < 5 && rc != CS_OK);
if (rc != CS_OK) {
crm_warn("Could not connect to Cluster Configuration Database API: %s (rc=%d)",
cs_strerror(rc), rc);
return -1;
}
rc = cmap_fd_get(cmap_handle, &fd);
if (rc != CS_OK) {
crm_err("Could not obtain the CMAP API connection: %s (%d)",
cs_strerror(rc), rc);
goto bail;
}
/* CMAP provider run as root (in given user namespace, anyway)? */
if (!(rv = crm_ipc_is_authentic_process(fd, (uid_t) 0,(gid_t) 0, &found_pid,
&found_uid, &found_gid))) {
crm_err("CMAP provider is not authentic:"
" process %lld (uid: %lld, gid: %lld)",
(long long) PCMK__SPECIAL_PID_AS_0(found_pid),
(long long) found_uid, (long long) found_gid);
goto bail;
} else if (rv < 0) {
crm_err("Could not verify authenticity of CMAP provider: %s (%d)",
strerror(-rv), -rv);
goto bail;
}
rc = cmap_iter_init(cmap_handle, prefix, &iter_handle);
if (rc != CS_OK) {
crm_warn("Failed to initialize iteration for corosync cmap '%s': %s (rc=%d)",
prefix, cs_strerror(rc), rc);
goto bail;
}
found = 0;
while ((rc = cmap_iter_next(cmap_handle, iter_handle, key_name, NULL, NULL)) == CS_OK) {
crm_trace("'%s' is configured in corosync cmap: %s", prefix, key_name);
found++;
break;
}
cmap_iter_finalize(cmap_handle, iter_handle);
bail:
cmap_finalize(cmap_handle);
return found;
}
diff --git a/lib/common/operations.c b/lib/common/operations.c
index e610bc00ea..ff6f3828ec 100644
--- a/lib/common/operations.c
+++ b/lib/common/operations.c
@@ -1,433 +1,431 @@
/*
* Copyright 2004-2020 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.
*/
#include <crm_internal.h>
#ifndef _GNU_SOURCE
# define _GNU_SOURCE
#endif
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <ctype.h>
#include <crm/crm.h>
#include <crm/lrmd.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/util.h>
/*!
* \brief Generate an operation key (RESOURCE_ACTION_INTERVAL)
*
* \param[in] rsc_id ID of resource being operated on
* \param[in] op_type Operation name
* \param[in] interval_ms Operation interval
*
* \return Newly allocated memory containing operation key as string
*
* \note This function asserts on errors, so it will never return NULL.
* The caller is responsible for freeing the result with free().
*/
char *
pcmk__op_key(const char *rsc_id, const char *op_type, guint interval_ms)
{
CRM_ASSERT(rsc_id != NULL);
CRM_ASSERT(op_type != NULL);
return crm_strdup_printf(PCMK__OP_FMT, rsc_id, op_type, interval_ms);
}
gboolean
parse_op_key(const char *key, char **rsc_id, char **op_type, guint *interval_ms)
{
char *notify = NULL;
char *mutable_key = NULL;
char *mutable_key_ptr = NULL;
size_t len = 0, offset = 0;
unsigned long long ch = 0;
guint local_interval_ms = 0;
// Initialize output variables in case of early return
if (rsc_id) {
*rsc_id = NULL;
}
if (op_type) {
*op_type = NULL;
}
if (interval_ms) {
*interval_ms = 0;
}
CRM_CHECK(key && *key, return FALSE);
// Parse interval at end of string
len = strlen(key);
offset = len - 1;
while ((offset > 0) && isdigit(key[offset])) {
ch = key[offset] - '0';
for (int digits = len - offset; digits > 1; --digits) {
ch = ch * 10;
}
local_interval_ms += ch;
offset--;
}
crm_trace("Operation key '%s' has interval %ums", key, local_interval_ms);
if (interval_ms) {
*interval_ms = local_interval_ms;
}
CRM_CHECK((offset != (len - 1)) && (key[offset] == '_'), return FALSE);
mutable_key = strndup(key, offset);
offset--;
while (offset > 0 && key[offset] != '_') {
offset--;
}
CRM_CHECK(key[offset] == '_',
free(mutable_key); return FALSE);
mutable_key_ptr = mutable_key + offset + 1;
crm_trace(" Action: %s", mutable_key_ptr);
if (op_type) {
*op_type = strdup(mutable_key_ptr);
}
mutable_key[offset] = 0;
offset--;
notify = strstr(mutable_key, "_post_notify");
if (notify && safe_str_eq(notify, "_post_notify")) {
notify[0] = 0;
}
notify = strstr(mutable_key, "_pre_notify");
if (notify && safe_str_eq(notify, "_pre_notify")) {
notify[0] = 0;
}
crm_trace(" Resource: %s", mutable_key);
if (rsc_id) {
*rsc_id = mutable_key;
} else {
free(mutable_key);
}
return TRUE;
}
char *
pcmk__notify_key(const char *rsc_id, const char *notify_type,
const char *op_type)
{
CRM_CHECK(rsc_id != NULL, return NULL);
CRM_CHECK(op_type != NULL, return NULL);
CRM_CHECK(notify_type != NULL, return NULL);
return crm_strdup_printf("%s_%s_notify_%s_0",
rsc_id, notify_type, op_type);
}
/*!
* \brief Parse a transition magic string into its constituent parts
*
* \param[in] magic Magic string to parse (must be non-NULL)
* \param[out] uuid If non-NULL, where to store copy of parsed UUID
* \param[out] transition_id If non-NULL, where to store parsed transition ID
* \param[out] action_id If non-NULL, where to store parsed action ID
* \param[out] op_status If non-NULL, where to store parsed result status
* \param[out] op_rc If non-NULL, where to store parsed actual rc
* \param[out] target_rc If non-NULL, where to stored parsed target rc
*
* \return TRUE if key was valid, FALSE otherwise
* \note If uuid is supplied and this returns TRUE, the caller is responsible
* for freeing the memory for *uuid using free().
*/
gboolean
decode_transition_magic(const char *magic, char **uuid, int *transition_id, int *action_id,
int *op_status, int *op_rc, int *target_rc)
{
int res = 0;
char *key = NULL;
gboolean result = TRUE;
int local_op_status = -1;
int local_op_rc = -1;
CRM_CHECK(magic != NULL, return FALSE);
#ifdef SSCANF_HAS_M
res = sscanf(magic, "%d:%d;%ms", &local_op_status, &local_op_rc, &key);
#else
key = calloc(1, strlen(magic) - 3); // magic must have >=4 other characters
CRM_ASSERT(key);
res = sscanf(magic, "%d:%d;%s", &local_op_status, &local_op_rc, key);
#endif
if (res == EOF) {
crm_err("Could not decode transition information '%s': %s",
magic, pcmk_strerror(errno));
result = FALSE;
} else if (res < 3) {
crm_warn("Transition information '%s' incomplete (%d of 3 expected items)",
magic, res);
result = FALSE;
} else {
if (op_status) {
*op_status = local_op_status;
}
if (op_rc) {
*op_rc = local_op_rc;
}
result = decode_transition_key(key, uuid, transition_id, action_id,
target_rc);
}
free(key);
return result;
}
char *
pcmk__transition_key(int transition_id, int action_id, int target_rc,
const char *node)
{
CRM_CHECK(node != NULL, return NULL);
return crm_strdup_printf("%d:%d:%d:%-*s",
action_id, transition_id, target_rc, 36, node);
}
/*!
* \brief Parse a transition key into its constituent parts
*
* \param[in] key Transition key to parse (must be non-NULL)
* \param[out] uuid If non-NULL, where to store copy of parsed UUID
* \param[out] transition_id If non-NULL, where to store parsed transition ID
* \param[out] action_id If non-NULL, where to store parsed action ID
* \param[out] target_rc If non-NULL, where to stored parsed target rc
*
* \return TRUE if key was valid, FALSE otherwise
* \note If uuid is supplied and this returns TRUE, the caller is responsible
* for freeing the memory for *uuid using free().
*/
gboolean
decode_transition_key(const char *key, char **uuid, int *transition_id, int *action_id,
int *target_rc)
{
int local_transition_id = -1;
int local_action_id = -1;
int local_target_rc = -1;
char local_uuid[37] = { '\0' };
// Initialize any supplied output arguments
if (uuid) {
*uuid = NULL;
}
if (transition_id) {
*transition_id = -1;
}
if (action_id) {
*action_id = -1;
}
if (target_rc) {
*target_rc = -1;
}
CRM_CHECK(key != NULL, return FALSE);
if (sscanf(key, "%d:%d:%d:%36s", &local_action_id, &local_transition_id,
&local_target_rc, local_uuid) != 4) {
crm_err("Invalid transition key '%s'", key);
return FALSE;
}
if (strlen(local_uuid) != 36) {
crm_warn("Invalid UUID '%s' in transition key '%s'", local_uuid, key);
}
if (uuid) {
*uuid = strdup(local_uuid);
CRM_ASSERT(*uuid);
}
if (transition_id) {
*transition_id = local_transition_id;
}
if (action_id) {
*action_id = local_action_id;
}
if (target_rc) {
*target_rc = local_target_rc;
}
return TRUE;
}
/*!
* \internal
* \brief Remove XML attributes not needed for operation digest
*
* \param[in,out] param_set XML with operation parameters
*/
void
pcmk__filter_op_for_digest(xmlNode *param_set)
{
char *key = NULL;
char *timeout = NULL;
guint interval_ms = 0;
const char *attr_filter[] = {
XML_ATTR_ID,
XML_ATTR_CRM_VERSION,
XML_LRM_ATTR_OP_DIGEST,
XML_LRM_ATTR_TARGET,
XML_LRM_ATTR_TARGET_UUID,
"pcmk_external_ip"
};
const int meta_len = strlen(CRM_META);
if (param_set == NULL) {
return;
}
// Remove the specific attributes listed in attr_filter
for (int lpc = 0; lpc < DIMOF(attr_filter); lpc++) {
xml_remove_prop(param_set, attr_filter[lpc]);
}
key = crm_meta_name(XML_LRM_ATTR_INTERVAL_MS);
if (crm_element_value_ms(param_set, key, &interval_ms) != pcmk_ok) {
interval_ms = 0;
}
free(key);
key = crm_meta_name(XML_ATTR_TIMEOUT);
timeout = crm_element_value_copy(param_set, key);
// Remove all CRM_meta_* attributes
for (xmlAttrPtr xIter = param_set->properties; xIter != NULL; ) {
const char *prop_name = (const char *) (xIter->name);
xIter = xIter->next;
// @TODO Why is this case-insensitive?
if (strncasecmp(prop_name, CRM_META, meta_len) == 0) {
xml_remove_prop(param_set, prop_name);
}
}
if ((interval_ms != 0) && (timeout != NULL)) {
// Add the timeout back, it's useful for recurring operation digests
crm_xml_add(param_set, key, timeout);
}
free(timeout);
free(key);
}
int
rsc_op_expected_rc(lrmd_event_data_t * op)
{
int rc = 0;
if (op && op->user_data) {
decode_transition_key(op->user_data, NULL, NULL, NULL, &rc);
}
return rc;
}
gboolean
did_rsc_op_fail(lrmd_event_data_t * op, int target_rc)
{
switch (op->op_status) {
case PCMK_LRM_OP_CANCELLED:
case PCMK_LRM_OP_PENDING:
return FALSE;
- break;
case PCMK_LRM_OP_NOTSUPPORTED:
case PCMK_LRM_OP_TIMEOUT:
case PCMK_LRM_OP_ERROR:
case PCMK_LRM_OP_NOT_CONNECTED:
case PCMK_LRM_OP_INVALID:
return TRUE;
- break;
default:
if (target_rc != op->rc) {
return TRUE;
}
}
return FALSE;
}
/*!
* \brief Create a CIB XML element for an operation
*
* \param[in] parent If not NULL, make new XML node a child of this one
* \param[in] prefix Generate an ID using this prefix
* \param[in] task Operation task to set
* \param[in] interval_spec Operation interval to set
* \param[in] timeout If not NULL, operation timeout to set
*
* \return New XML object on success, NULL otherwise
*/
xmlNode *
crm_create_op_xml(xmlNode *parent, const char *prefix, const char *task,
const char *interval_spec, const char *timeout)
{
xmlNode *xml_op;
CRM_CHECK(prefix && task && interval_spec, return NULL);
xml_op = create_xml_node(parent, XML_ATTR_OP);
crm_xml_set_id(xml_op, "%s-%s-%s", prefix, task, interval_spec);
crm_xml_add(xml_op, XML_LRM_ATTR_INTERVAL, interval_spec);
crm_xml_add(xml_op, "name", task);
if (timeout) {
crm_xml_add(xml_op, XML_ATTR_TIMEOUT, timeout);
}
return xml_op;
}
/*!
* \brief Check whether an operation requires resource agent meta-data
*
* \param[in] rsc_class Resource agent class (or NULL to skip class check)
* \param[in] op Operation action (or NULL to skip op check)
*
* \return TRUE if operation needs meta-data, FALSE otherwise
* \note At least one of rsc_class and op must be specified.
*/
bool
crm_op_needs_metadata(const char *rsc_class, const char *op)
{
/* Agent meta-data is used to determine whether a reload is possible, and to
* evaluate versioned parameters -- so if this op is not relevant to those
* features, we don't need the meta-data.
*/
CRM_CHECK(rsc_class || op, return FALSE);
if (rsc_class
&& is_not_set(pcmk_get_ra_caps(rsc_class), pcmk_ra_cap_params)) {
/* Meta-data is only needed for resource classes that use parameters */
return FALSE;
}
/* Meta-data is only needed for these actions */
if (op
&& strcmp(op, CRMD_ACTION_START)
&& strcmp(op, CRMD_ACTION_STATUS)
&& strcmp(op, CRMD_ACTION_PROMOTE)
&& strcmp(op, CRMD_ACTION_DEMOTE)
&& strcmp(op, CRMD_ACTION_RELOAD)
&& strcmp(op, CRMD_ACTION_MIGRATE)
&& strcmp(op, CRMD_ACTION_MIGRATED)
&& strcmp(op, CRMD_ACTION_NOTIFY)) {
return FALSE;
}
return TRUE;
}
diff --git a/lib/lrmd/lrmd_client.c b/lib/lrmd/lrmd_client.c
index a6c023b9bb..e048c082d8 100644
--- a/lib/lrmd/lrmd_client.c
+++ b/lib/lrmd/lrmd_client.c
@@ -1,2103 +1,2101 @@
/*
* Copyright 2012-2020 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.
*/
#include <crm_internal.h>
#include <unistd.h>
#include <stdlib.h>
#include <stdio.h>
#include <stdarg.h>
#include <string.h>
#include <ctype.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <glib.h>
#include <dirent.h>
#include <crm/crm.h>
#include <crm/lrmd.h>
#include <crm/lrmd_internal.h>
#include <crm/services.h>
#include <crm/common/mainloop.h>
#include <crm/common/ipc_internal.h>
#include <crm/common/remote_internal.h>
#include <crm/msg_xml.h>
#include <crm/stonith-ng.h>
#ifdef HAVE_GNUTLS_GNUTLS_H
# undef KEYFILE
# include <gnutls/gnutls.h>
#endif
#include <sys/socket.h>
#include <netinet/in.h>
#include <netinet/ip.h>
#include <arpa/inet.h>
#include <netdb.h>
#define MAX_TLS_RECV_WAIT 10000
CRM_TRACE_INIT_DATA(lrmd);
static int lrmd_api_disconnect(lrmd_t * lrmd);
static int lrmd_api_is_connected(lrmd_t * lrmd);
/* IPC proxy functions */
int lrmd_internal_proxy_send(lrmd_t * lrmd, xmlNode *msg);
static void lrmd_internal_proxy_dispatch(lrmd_t *lrmd, xmlNode *msg);
void lrmd_internal_set_proxy_callback(lrmd_t * lrmd, void *userdata, void (*callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg));
#ifdef HAVE_GNUTLS_GNUTLS_H
# define LRMD_CLIENT_HANDSHAKE_TIMEOUT 5000 /* 5 seconds */
gnutls_psk_client_credentials_t psk_cred_s;
int lrmd_tls_set_key(gnutls_datum_t * key);
static void lrmd_tls_disconnect(lrmd_t * lrmd);
static int global_remote_msg_id = 0;
static void lrmd_tls_connection_destroy(gpointer userdata);
#endif
typedef struct lrmd_private_s {
enum pcmk__client_type type;
char *token;
mainloop_io_t *source;
/* IPC parameters */
crm_ipc_t *ipc;
pcmk__remote_t *remote;
/* Extra TLS parameters */
char *remote_nodename;
#ifdef HAVE_GNUTLS_GNUTLS_H
char *server;
int port;
gnutls_psk_client_credentials_t psk_cred_c;
/* while the async connection is occurring, this is the id
* of the connection timeout timer. */
int async_timer;
int sock;
/* since tls requires a round trip across the network for a
* request/reply, there are times where we just want to be able
* to send a request from the client and not wait around (or even care
* about) what the reply is. */
int expected_late_replies;
GList *pending_notify;
crm_trigger_t *process_notify;
#endif
lrmd_event_callback callback;
/* Internal IPC proxy msg passing for remote guests */
void (*proxy_callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg);
void *proxy_callback_userdata;
char *peer_version;
} lrmd_private_t;
static lrmd_list_t *
lrmd_list_add(lrmd_list_t * head, const char *value)
{
lrmd_list_t *p, *end;
p = calloc(1, sizeof(lrmd_list_t));
p->val = strdup(value);
end = head;
while (end && end->next) {
end = end->next;
}
if (end) {
end->next = p;
} else {
head = p;
}
return head;
}
void
lrmd_list_freeall(lrmd_list_t * head)
{
lrmd_list_t *p;
while (head) {
char *val = (char *)head->val;
p = head->next;
free(val);
free(head);
head = p;
}
}
lrmd_key_value_t *
lrmd_key_value_add(lrmd_key_value_t * head, const char *key, const char *value)
{
lrmd_key_value_t *p, *end;
p = calloc(1, sizeof(lrmd_key_value_t));
p->key = strdup(key);
p->value = strdup(value);
end = head;
while (end && end->next) {
end = end->next;
}
if (end) {
end->next = p;
} else {
head = p;
}
return head;
}
void
lrmd_key_value_freeall(lrmd_key_value_t * head)
{
lrmd_key_value_t *p;
while (head) {
p = head->next;
free(head->key);
free(head->value);
free(head);
head = p;
}
}
/*!
* Create a new lrmd_event_data_t object
*
* \param[in] rsc_id ID of resource involved in event
* \param[in] task Action name
* \param[in] interval_ms Action interval
*
* \return Newly allocated and initialized lrmd_event_data_t
* \note This functions asserts on memory errors, so the return value is
* guaranteed to be non-NULL. The caller is responsible for freeing the
* result with lrmd_free_event().
*/
lrmd_event_data_t *
lrmd_new_event(const char *rsc_id, const char *task, guint interval_ms)
{
lrmd_event_data_t *event = calloc(1, sizeof(lrmd_event_data_t));
CRM_ASSERT(event != NULL);
if (rsc_id != NULL) {
event->rsc_id = strdup(rsc_id);
CRM_ASSERT(event->rsc_id != NULL);
}
if (task != NULL) {
event->op_type = strdup(task);
CRM_ASSERT(event->op_type != NULL);
}
event->interval_ms = interval_ms;
return event;
}
lrmd_event_data_t *
lrmd_copy_event(lrmd_event_data_t * event)
{
lrmd_event_data_t *copy = NULL;
copy = calloc(1, sizeof(lrmd_event_data_t));
/* This will get all the int values.
* we just have to be careful not to leave any
* dangling pointers to strings. */
memcpy(copy, event, sizeof(lrmd_event_data_t));
copy->rsc_id = event->rsc_id ? strdup(event->rsc_id) : NULL;
copy->op_type = event->op_type ? strdup(event->op_type) : NULL;
copy->user_data = event->user_data ? strdup(event->user_data) : NULL;
copy->output = event->output ? strdup(event->output) : NULL;
copy->exit_reason = event->exit_reason ? strdup(event->exit_reason) : NULL;
copy->remote_nodename = event->remote_nodename ? strdup(event->remote_nodename) : NULL;
copy->params = crm_str_table_dup(event->params);
return copy;
}
void
lrmd_free_event(lrmd_event_data_t * event)
{
if (!event) {
return;
}
/* free gives me grief if i try to cast */
free((char *)event->rsc_id);
free((char *)event->op_type);
free((char *)event->user_data);
free((char *)event->output);
free((char *)event->exit_reason);
free((char *)event->remote_nodename);
if (event->params) {
g_hash_table_destroy(event->params);
}
free(event);
}
static int
lrmd_dispatch_internal(lrmd_t * lrmd, xmlNode * msg)
{
const char *type;
const char *proxy_session = crm_element_value(msg, F_LRMD_IPC_SESSION);
lrmd_private_t *native = lrmd->lrmd_private;
lrmd_event_data_t event = { 0, };
if (proxy_session != NULL) {
/* this is proxy business */
lrmd_internal_proxy_dispatch(lrmd, msg);
return 1;
} else if (!native->callback) {
/* no callback set */
crm_trace("notify event received but client has not set callback");
return 1;
}
event.remote_nodename = native->remote_nodename;
type = crm_element_value(msg, F_LRMD_OPERATION);
crm_element_value_int(msg, F_LRMD_CALLID, &event.call_id);
event.rsc_id = crm_element_value(msg, F_LRMD_RSC_ID);
if (crm_str_eq(type, LRMD_OP_RSC_REG, TRUE)) {
event.type = lrmd_event_register;
} else if (crm_str_eq(type, LRMD_OP_RSC_UNREG, TRUE)) {
event.type = lrmd_event_unregister;
} else if (crm_str_eq(type, LRMD_OP_RSC_EXEC, TRUE)) {
time_t epoch = 0;
crm_element_value_int(msg, F_LRMD_TIMEOUT, &event.timeout);
crm_element_value_ms(msg, F_LRMD_RSC_INTERVAL, &event.interval_ms);
crm_element_value_int(msg, F_LRMD_RSC_START_DELAY, &event.start_delay);
crm_element_value_int(msg, F_LRMD_EXEC_RC, (int *)&event.rc);
crm_element_value_int(msg, F_LRMD_OP_STATUS, &event.op_status);
crm_element_value_int(msg, F_LRMD_RSC_DELETED, &event.rsc_deleted);
crm_element_value_epoch(msg, F_LRMD_RSC_RUN_TIME, &epoch);
event.t_run = (unsigned int) epoch;
crm_element_value_epoch(msg, F_LRMD_RSC_RCCHANGE_TIME, &epoch);
event.t_rcchange = (unsigned int) epoch;
crm_element_value_int(msg, F_LRMD_RSC_EXEC_TIME, (int *)&event.exec_time);
crm_element_value_int(msg, F_LRMD_RSC_QUEUE_TIME, (int *)&event.queue_time);
event.op_type = crm_element_value(msg, F_LRMD_RSC_ACTION);
event.user_data = crm_element_value(msg, F_LRMD_RSC_USERDATA_STR);
event.output = crm_element_value(msg, F_LRMD_RSC_OUTPUT);
event.exit_reason = crm_element_value(msg, F_LRMD_RSC_EXIT_REASON);
event.type = lrmd_event_exec_complete;
event.params = xml2list(msg);
} else if (crm_str_eq(type, LRMD_OP_NEW_CLIENT, TRUE)) {
event.type = lrmd_event_new_client;
} else if (crm_str_eq(type, LRMD_OP_POKE, TRUE)) {
event.type = lrmd_event_poke;
} else {
return 1;
}
crm_trace("op %s notify event received", type);
native->callback(&event);
if (event.params) {
g_hash_table_destroy(event.params);
}
return 1;
}
static int
lrmd_ipc_dispatch(const char *buffer, ssize_t length, gpointer userdata)
{
lrmd_t *lrmd = userdata;
lrmd_private_t *native = lrmd->lrmd_private;
xmlNode *msg;
int rc;
if (!native->callback) {
/* no callback set */
return 1;
}
msg = string2xml(buffer);
rc = lrmd_dispatch_internal(lrmd, msg);
free_xml(msg);
return rc;
}
#ifdef HAVE_GNUTLS_GNUTLS_H
static void
lrmd_free_xml(gpointer userdata)
{
free_xml((xmlNode *) userdata);
}
static int
lrmd_tls_connected(lrmd_t * lrmd)
{
lrmd_private_t *native = lrmd->lrmd_private;
if (native->remote->tls_session) {
return TRUE;
}
return FALSE;
}
static int
lrmd_tls_dispatch(gpointer userdata)
{
lrmd_t *lrmd = userdata;
lrmd_private_t *native = lrmd->lrmd_private;
xmlNode *xml = NULL;
int rc = pcmk_rc_ok;
if (lrmd_tls_connected(lrmd) == FALSE) {
crm_trace("TLS dispatch triggered after disconnect");
return 0;
}
crm_trace("TLS dispatch triggered");
/* First check if there are any pending notifies to process that came
* while we were waiting for replies earlier. */
if (native->pending_notify) {
GList *iter = NULL;
crm_trace("Processing pending notifies");
for (iter = native->pending_notify; iter; iter = iter->next) {
lrmd_dispatch_internal(lrmd, iter->data);
}
g_list_free_full(native->pending_notify, lrmd_free_xml);
native->pending_notify = NULL;
}
/* Next read the current buffer and see if there are any messages to handle. */
switch (pcmk__remote_ready(native->remote, 0)) {
case pcmk_rc_ok:
rc = pcmk__read_remote_message(native->remote, -1);
xml = pcmk__remote_message_xml(native->remote);
break;
case ETIME:
// Nothing to read, check if a full message is already in buffer
xml = pcmk__remote_message_xml(native->remote);
break;
default:
rc = ENOTCONN;
break;
}
while (xml) {
const char *msg_type = crm_element_value(xml, F_LRMD_REMOTE_MSG_TYPE);
if (safe_str_eq(msg_type, "notify")) {
lrmd_dispatch_internal(lrmd, xml);
} else if (safe_str_eq(msg_type, "reply")) {
if (native->expected_late_replies > 0) {
native->expected_late_replies--;
} else {
int reply_id = 0;
crm_element_value_int(xml, F_LRMD_CALLID, &reply_id);
/* if this happens, we want to know about it */
crm_err("Got outdated Pacemaker Remote reply %d", reply_id);
}
}
free_xml(xml);
xml = pcmk__remote_message_xml(native->remote);
}
if (rc == ENOTCONN) {
crm_info("Lost %s executor connection while reading data",
(native->remote_nodename? native->remote_nodename : "local"));
lrmd_tls_disconnect(lrmd);
return 0;
}
return 1;
}
#endif
/* Not used with mainloop */
int
lrmd_poll(lrmd_t * lrmd, int timeout)
{
lrmd_private_t *native = lrmd->lrmd_private;
switch (native->type) {
case PCMK__CLIENT_IPC:
return crm_ipc_ready(native->ipc);
#ifdef HAVE_GNUTLS_GNUTLS_H
case PCMK__CLIENT_TLS:
if (native->pending_notify) {
return 1;
} else {
int rc = pcmk__remote_ready(native->remote, 0);
switch (rc) {
case pcmk_rc_ok:
return 1;
case ETIME:
return 0;
default:
return pcmk_rc2legacy(rc);
}
}
#endif
default:
crm_err("Unsupported connection type: %d", native->type);
}
return 0;
}
/* Not used with mainloop */
bool
lrmd_dispatch(lrmd_t * lrmd)
{
lrmd_private_t *private = NULL;
CRM_ASSERT(lrmd != NULL);
private = lrmd->lrmd_private;
switch (private->type) {
case PCMK__CLIENT_IPC:
while (crm_ipc_ready(private->ipc)) {
if (crm_ipc_read(private->ipc) > 0) {
const char *msg = crm_ipc_buffer(private->ipc);
lrmd_ipc_dispatch(msg, strlen(msg), lrmd);
}
}
break;
#ifdef HAVE_GNUTLS_GNUTLS_H
case PCMK__CLIENT_TLS:
lrmd_tls_dispatch(lrmd);
break;
#endif
default:
crm_err("Unsupported connection type: %d", private->type);
}
if (lrmd_api_is_connected(lrmd) == FALSE) {
crm_err("Connection closed");
return FALSE;
}
return TRUE;
}
static xmlNode *
lrmd_create_op(const char *token, const char *op, xmlNode *data, int timeout,
enum lrmd_call_options options)
{
xmlNode *op_msg = create_xml_node(NULL, "lrmd_command");
CRM_CHECK(op_msg != NULL, return NULL);
CRM_CHECK(token != NULL, return NULL);
crm_xml_add(op_msg, F_XML_TAGNAME, "lrmd_command");
crm_xml_add(op_msg, F_TYPE, T_LRMD);
crm_xml_add(op_msg, F_LRMD_CALLBACK_TOKEN, token);
crm_xml_add(op_msg, F_LRMD_OPERATION, op);
crm_xml_add_int(op_msg, F_LRMD_TIMEOUT, timeout);
crm_xml_add_int(op_msg, F_LRMD_CALLOPTS, options);
if (data != NULL) {
add_message_xml(op_msg, F_LRMD_CALLDATA, data);
}
crm_trace("Created executor %s command with call options %.8lx (%d)",
op, (long)options, options);
return op_msg;
}
static void
lrmd_ipc_connection_destroy(gpointer userdata)
{
lrmd_t *lrmd = userdata;
lrmd_private_t *native = lrmd->lrmd_private;
crm_info("IPC connection destroyed");
/* Prevent these from being cleaned up in lrmd_api_disconnect() */
native->ipc = NULL;
native->source = NULL;
if (native->callback) {
lrmd_event_data_t event = { 0, };
event.type = lrmd_event_disconnect;
event.remote_nodename = native->remote_nodename;
native->callback(&event);
}
}
#ifdef HAVE_GNUTLS_GNUTLS_H
static void
lrmd_tls_connection_destroy(gpointer userdata)
{
lrmd_t *lrmd = userdata;
lrmd_private_t *native = lrmd->lrmd_private;
crm_info("TLS connection destroyed");
if (native->remote->tls_session) {
gnutls_bye(*native->remote->tls_session, GNUTLS_SHUT_RDWR);
gnutls_deinit(*native->remote->tls_session);
gnutls_free(native->remote->tls_session);
}
if (native->psk_cred_c) {
gnutls_psk_free_client_credentials(native->psk_cred_c);
}
if (native->sock) {
close(native->sock);
}
if (native->process_notify) {
mainloop_destroy_trigger(native->process_notify);
native->process_notify = NULL;
}
if (native->pending_notify) {
g_list_free_full(native->pending_notify, lrmd_free_xml);
native->pending_notify = NULL;
}
free(native->remote->buffer);
native->remote->buffer = NULL;
native->source = 0;
native->sock = 0;
native->psk_cred_c = NULL;
native->remote->tls_session = NULL;
native->sock = 0;
if (native->callback) {
lrmd_event_data_t event = { 0, };
event.remote_nodename = native->remote_nodename;
event.type = lrmd_event_disconnect;
native->callback(&event);
}
return;
}
// \return Standard Pacemaker return code
int
lrmd_tls_send_msg(pcmk__remote_t *session, xmlNode *msg, uint32_t id,
const char *msg_type)
{
crm_xml_add_int(msg, F_LRMD_REMOTE_MSG_ID, id);
crm_xml_add(msg, F_LRMD_REMOTE_MSG_TYPE, msg_type);
return pcmk__remote_send_xml(session, msg);
}
static xmlNode *
lrmd_tls_recv_reply(lrmd_t * lrmd, int total_timeout, int expected_reply_id, int *disconnected)
{
lrmd_private_t *native = lrmd->lrmd_private;
xmlNode *xml = NULL;
time_t start = time(NULL);
const char *msg_type = NULL;
int reply_id = 0;
int remaining_timeout = 0;
/* A timeout of 0 here makes no sense. We have to wait a period of time
* for the response to come back. If -1 or 0, default to 10 seconds. */
if (total_timeout <= 0 || total_timeout > MAX_TLS_RECV_WAIT) {
total_timeout = MAX_TLS_RECV_WAIT;
}
while (!xml) {
xml = pcmk__remote_message_xml(native->remote);
if (!xml) {
/* read some more off the tls buffer if we still have time left. */
if (remaining_timeout) {
remaining_timeout = total_timeout - ((time(NULL) - start) * 1000);
} else {
remaining_timeout = total_timeout;
}
if (remaining_timeout <= 0) {
crm_err("Never received the expected reply during the timeout period, disconnecting.");
*disconnected = TRUE;
return NULL;
}
if (pcmk__read_remote_message(native->remote,
remaining_timeout) == ENOTCONN) {
*disconnected = TRUE;
} else {
*disconnected = FALSE;
}
xml = pcmk__remote_message_xml(native->remote);
if (!xml) {
crm_err("Unable to receive expected reply, disconnecting.");
*disconnected = TRUE;
return NULL;
} else if (*disconnected) {
return NULL;
}
}
CRM_ASSERT(xml != NULL);
crm_element_value_int(xml, F_LRMD_REMOTE_MSG_ID, &reply_id);
msg_type = crm_element_value(xml, F_LRMD_REMOTE_MSG_TYPE);
if (!msg_type) {
crm_err("Empty msg type received while waiting for reply");
free_xml(xml);
xml = NULL;
} else if (safe_str_eq(msg_type, "notify")) {
/* got a notify while waiting for reply, trigger the notify to be processed later */
crm_info("queueing notify");
native->pending_notify = g_list_append(native->pending_notify, xml);
if (native->process_notify) {
crm_info("notify trigger set.");
mainloop_set_trigger(native->process_notify);
}
xml = NULL;
} else if (safe_str_neq(msg_type, "reply")) {
/* msg isn't a reply, make some noise */
crm_err("Expected a reply, got %s", msg_type);
free_xml(xml);
xml = NULL;
} else if (reply_id != expected_reply_id) {
if (native->expected_late_replies > 0) {
native->expected_late_replies--;
} else {
crm_err("Got outdated reply, expected id %d got id %d", expected_reply_id, reply_id);
}
free_xml(xml);
xml = NULL;
}
}
if (native->remote->buffer && native->process_notify) {
mainloop_set_trigger(native->process_notify);
}
return xml;
}
static int
lrmd_tls_send(lrmd_t * lrmd, xmlNode * msg)
{
int rc = 0;
lrmd_private_t *native = lrmd->lrmd_private;
global_remote_msg_id++;
if (global_remote_msg_id <= 0) {
global_remote_msg_id = 1;
}
rc = lrmd_tls_send_msg(native->remote, msg, global_remote_msg_id, "request");
if (rc != pcmk_rc_ok) {
crm_err("Disconnecting because TLS message could not be sent to "
"Pacemaker Remote: %s", pcmk_rc_str(rc));
lrmd_tls_disconnect(lrmd);
return -ENOTCONN;
}
return pcmk_ok;
}
static int
lrmd_tls_send_recv(lrmd_t * lrmd, xmlNode * msg, int timeout, xmlNode ** reply)
{
int rc = 0;
int disconnected = 0;
xmlNode *xml = NULL;
if (lrmd_tls_connected(lrmd) == FALSE) {
return -1;
}
rc = lrmd_tls_send(lrmd, msg);
if (rc < 0) {
return rc;
}
xml = lrmd_tls_recv_reply(lrmd, timeout, global_remote_msg_id, &disconnected);
if (disconnected) {
crm_err("Pacemaker Remote disconnected while waiting for reply to request id %d",
global_remote_msg_id);
lrmd_tls_disconnect(lrmd);
rc = -ENOTCONN;
} else if (!xml) {
crm_err("Did not receive reply from Pacemaker Remote for request id %d (timeout %dms)",
global_remote_msg_id, timeout);
rc = -ECOMM;
}
if (reply) {
*reply = xml;
} else {
free_xml(xml);
}
return rc;
}
#endif
static int
lrmd_send_xml(lrmd_t * lrmd, xmlNode * msg, int timeout, xmlNode ** reply)
{
int rc = -1;
lrmd_private_t *native = lrmd->lrmd_private;
switch (native->type) {
case PCMK__CLIENT_IPC:
rc = crm_ipc_send(native->ipc, msg, crm_ipc_client_response, timeout, reply);
break;
#ifdef HAVE_GNUTLS_GNUTLS_H
case PCMK__CLIENT_TLS:
rc = lrmd_tls_send_recv(lrmd, msg, timeout, reply);
break;
#endif
default:
crm_err("Unsupported connection type: %d", native->type);
}
return rc;
}
static int
lrmd_send_xml_no_reply(lrmd_t * lrmd, xmlNode * msg)
{
int rc = -1;
lrmd_private_t *native = lrmd->lrmd_private;
switch (native->type) {
case PCMK__CLIENT_IPC:
rc = crm_ipc_send(native->ipc, msg, crm_ipc_flags_none, 0, NULL);
break;
#ifdef HAVE_GNUTLS_GNUTLS_H
case PCMK__CLIENT_TLS:
rc = lrmd_tls_send(lrmd, msg);
if (rc == pcmk_ok) {
/* we don't want to wait around for the reply, but
* since the request/reply protocol needs to behave the same
* as libqb, a reply will eventually come later anyway. */
native->expected_late_replies++;
}
break;
#endif
default:
crm_err("Unsupported connection type: %d", native->type);
}
return rc;
}
static int
lrmd_api_is_connected(lrmd_t * lrmd)
{
lrmd_private_t *native = lrmd->lrmd_private;
switch (native->type) {
case PCMK__CLIENT_IPC:
return crm_ipc_connected(native->ipc);
- break;
#ifdef HAVE_GNUTLS_GNUTLS_H
case PCMK__CLIENT_TLS:
return lrmd_tls_connected(lrmd);
- break;
#endif
default:
crm_err("Unsupported connection type: %d", native->type);
}
return 0;
}
/*!
* \internal
* \brief Send a prepared API command to the executor
*
* \param[in] lrmd Existing connection to the executor
* \param[in] op Name of API command to send
* \param[in] data Command data XML to add to the sent command
* \param[out] output_data If expecting a reply, it will be stored here
* \param[in] timeout Timeout in milliseconds (if 0, defaults to
* a sensible value per the type of connection,
* standard vs. pacemaker remote);
* also propagated to the command XML
* \param[in] call_options Call options to pass to server when sending
* \param[in] expect_reply If TRUE, wait for a reply from the server;
* must be TRUE for IPC (as opposed to TLS) clients
*
* \return pcmk_ok on success, -errno on error
*/
static int
lrmd_send_command(lrmd_t *lrmd, const char *op, xmlNode *data,
xmlNode **output_data, int timeout,
enum lrmd_call_options options, gboolean expect_reply)
{
int rc = pcmk_ok;
lrmd_private_t *native = lrmd->lrmd_private;
xmlNode *op_msg = NULL;
xmlNode *op_reply = NULL;
if (!lrmd_api_is_connected(lrmd)) {
return -ENOTCONN;
}
if (op == NULL) {
crm_err("No operation specified");
return -EINVAL;
}
CRM_CHECK(native->token != NULL,;
);
crm_trace("Sending %s op to executor", op);
op_msg = lrmd_create_op(native->token, op, data, timeout, options);
if (op_msg == NULL) {
return -EINVAL;
}
if (expect_reply) {
rc = lrmd_send_xml(lrmd, op_msg, timeout, &op_reply);
} else {
rc = lrmd_send_xml_no_reply(lrmd, op_msg);
goto done;
}
if (rc < 0) {
crm_perror(LOG_ERR, "Couldn't perform %s operation (timeout=%d): %d", op, timeout, rc);
rc = -ECOMM;
goto done;
} else if(op_reply == NULL) {
rc = -ENOMSG;
goto done;
}
rc = pcmk_ok;
crm_trace("%s op reply received", op);
if (crm_element_value_int(op_reply, F_LRMD_RC, &rc) != 0) {
rc = -ENOMSG;
goto done;
}
crm_log_xml_trace(op_reply, "Reply");
if (output_data) {
*output_data = op_reply;
op_reply = NULL; /* Prevent subsequent free */
}
done:
if (lrmd_api_is_connected(lrmd) == FALSE) {
crm_err("Executor disconnected");
}
free_xml(op_msg);
free_xml(op_reply);
return rc;
}
static int
lrmd_api_poke_connection(lrmd_t * lrmd)
{
int rc;
lrmd_private_t *native = lrmd->lrmd_private;
xmlNode *data = create_xml_node(NULL, F_LRMD_RSC);
crm_xml_add(data, F_LRMD_ORIGIN, __FUNCTION__);
rc = lrmd_send_command(lrmd, LRMD_OP_POKE, data, NULL, 0, 0,
(native->type == PCMK__CLIENT_IPC));
free_xml(data);
return rc < 0 ? rc : pcmk_ok;
}
int
remote_proxy_check(lrmd_t * lrmd, GHashTable *hash)
{
int rc;
const char *value;
lrmd_private_t *native = lrmd->lrmd_private;
xmlNode *data = create_xml_node(NULL, F_LRMD_OPERATION);
crm_xml_add(data, F_LRMD_ORIGIN, __FUNCTION__);
value = g_hash_table_lookup(hash, "stonith-watchdog-timeout");
crm_xml_add(data, F_LRMD_WATCHDOG, value);
rc = lrmd_send_command(lrmd, LRMD_OP_CHECK, data, NULL, 0, 0,
(native->type == PCMK__CLIENT_IPC));
free_xml(data);
return rc < 0 ? rc : pcmk_ok;
}
static int
lrmd_handshake(lrmd_t * lrmd, const char *name)
{
int rc = pcmk_ok;
lrmd_private_t *native = lrmd->lrmd_private;
xmlNode *reply = NULL;
xmlNode *hello = create_xml_node(NULL, "lrmd_command");
crm_xml_add(hello, F_TYPE, T_LRMD);
crm_xml_add(hello, F_LRMD_OPERATION, CRM_OP_REGISTER);
crm_xml_add(hello, F_LRMD_CLIENTNAME, name);
crm_xml_add(hello, F_LRMD_PROTOCOL_VERSION, LRMD_PROTOCOL_VERSION);
/* advertise that we are a proxy provider */
if (native->proxy_callback) {
crm_xml_add(hello, F_LRMD_IS_IPC_PROVIDER, "true");
}
rc = lrmd_send_xml(lrmd, hello, -1, &reply);
if (rc < 0) {
crm_perror(LOG_DEBUG, "Couldn't complete registration with the executor API: %d", rc);
rc = -ECOMM;
} else if (reply == NULL) {
crm_err("Did not receive registration reply");
rc = -EPROTO;
} else {
const char *version = crm_element_value(reply, F_LRMD_PROTOCOL_VERSION);
const char *msg_type = crm_element_value(reply, F_LRMD_OPERATION);
const char *tmp_ticket = crm_element_value(reply, F_LRMD_CLIENTID);
crm_element_value_int(reply, F_LRMD_RC, &rc);
if (rc == -EPROTO) {
crm_err("Executor protocol version mismatch between client (%s) and server (%s)",
LRMD_PROTOCOL_VERSION, version);
crm_log_xml_err(reply, "Protocol Error");
} else if (safe_str_neq(msg_type, CRM_OP_REGISTER)) {
crm_err("Invalid registration message: %s", msg_type);
crm_log_xml_err(reply, "Bad reply");
rc = -EPROTO;
} else if (tmp_ticket == NULL) {
crm_err("No registration token provided");
crm_log_xml_err(reply, "Bad reply");
rc = -EPROTO;
} else {
crm_trace("Obtained registration token: %s", tmp_ticket);
native->token = strdup(tmp_ticket);
native->peer_version = strdup(version?version:"1.0"); /* Included since 1.1 */
rc = pcmk_ok;
}
}
free_xml(reply);
free_xml(hello);
if (rc != pcmk_ok) {
lrmd_api_disconnect(lrmd);
}
return rc;
}
static int
lrmd_ipc_connect(lrmd_t * lrmd, int *fd)
{
int rc = pcmk_ok;
lrmd_private_t *native = lrmd->lrmd_private;
struct ipc_client_callbacks lrmd_callbacks = {
.dispatch = lrmd_ipc_dispatch,
.destroy = lrmd_ipc_connection_destroy
};
crm_info("Connecting to executor");
if (fd) {
/* No mainloop */
native->ipc = crm_ipc_new(CRM_SYSTEM_LRMD, 0);
if (native->ipc && crm_ipc_connect(native->ipc)) {
*fd = crm_ipc_get_fd(native->ipc);
} else if (native->ipc) {
crm_perror(LOG_ERR, "Connection to executor failed");
rc = -ENOTCONN;
}
} else {
native->source = mainloop_add_ipc_client(CRM_SYSTEM_LRMD, G_PRIORITY_HIGH, 0, lrmd, &lrmd_callbacks);
native->ipc = mainloop_get_ipc_client(native->source);
}
if (native->ipc == NULL) {
crm_debug("Could not connect to the executor API");
rc = -ENOTCONN;
}
return rc;
}
#ifdef HAVE_GNUTLS_GNUTLS_H
static void
copy_gnutls_datum(gnutls_datum_t *dest, gnutls_datum_t *source)
{
dest->data = gnutls_malloc(source->size);
CRM_ASSERT(dest->data);
memcpy(dest->data, source->data, source->size);
dest->size = source->size;
}
static void
clear_gnutls_datum(gnutls_datum_t *datum)
{
gnutls_free(datum->data);
datum->data = NULL;
datum->size = 0;
}
#define KEY_READ_LEN 256
static int
set_key(gnutls_datum_t * key, const char *location)
{
FILE *stream;
size_t buf_len = KEY_READ_LEN;
static gnutls_datum_t key_cache = { 0, };
static time_t key_cache_updated = 0;
if (location == NULL) {
return -1;
}
if (key_cache.data != NULL) {
if ((time(NULL) - key_cache_updated) < 60) {
copy_gnutls_datum(key, &key_cache);
crm_debug("Using cached Pacemaker Remote key");
return 0;
} else {
clear_gnutls_datum(&key_cache);
key_cache_updated = 0;
crm_debug("Cleared Pacemaker Remote key cache");
}
}
stream = fopen(location, "r");
if (!stream) {
return -1;
}
key->data = gnutls_malloc(buf_len);
key->size = 0;
while (!feof(stream)) {
int next = fgetc(stream);
if (next == EOF) {
if (!feof(stream)) {
crm_err("Error reading Pacemaker Remote key; copy in memory may be corrupted");
}
break;
}
if (key->size == buf_len) {
buf_len = key->size + KEY_READ_LEN;
key->data = gnutls_realloc(key->data, buf_len);
CRM_ASSERT(key->data);
}
key->data[key->size++] = (unsigned char) next;
}
fclose(stream);
if (key->size == 0) {
clear_gnutls_datum(key);
return -1;
}
if (key_cache.data == NULL) {
copy_gnutls_datum(&key_cache, key);
key_cache_updated = time(NULL);
crm_debug("Cached Pacemaker Remote key");
}
return 0;
}
int
lrmd_tls_set_key(gnutls_datum_t * key)
{
const char *specific_location = getenv("PCMK_authkey_location");
if (set_key(key, specific_location) == 0) {
crm_debug("Using custom authkey location %s", specific_location);
return pcmk_ok;
} else if (specific_location) {
crm_err("No valid Pacemaker Remote key found at %s, trying default location", specific_location);
}
if ((set_key(key, DEFAULT_REMOTE_KEY_LOCATION) != 0)
&& (set_key(key, ALT_REMOTE_KEY_LOCATION) != 0)) {
crm_err("No valid Pacemaker Remote key found at %s", DEFAULT_REMOTE_KEY_LOCATION);
return -ENOKEY;
}
return pcmk_ok;
}
static void
lrmd_gnutls_global_init(void)
{
static int gnutls_init = 0;
if (!gnutls_init) {
crm_gnutls_global_init();
}
gnutls_init = 1;
}
#endif
static void
report_async_connection_result(lrmd_t * lrmd, int rc)
{
lrmd_private_t *native = lrmd->lrmd_private;
if (native->callback) {
lrmd_event_data_t event = { 0, };
event.type = lrmd_event_connect;
event.remote_nodename = native->remote_nodename;
event.connection_rc = rc;
native->callback(&event);
}
}
#ifdef HAVE_GNUTLS_GNUTLS_H
static inline int
lrmd__tls_client_handshake(pcmk__remote_t *remote)
{
return pcmk__tls_client_handshake(remote, LRMD_CLIENT_HANDSHAKE_TIMEOUT);
}
static void
lrmd_tcp_connect_cb(void *userdata, int rc, int sock)
{
lrmd_t *lrmd = userdata;
lrmd_private_t *native = lrmd->lrmd_private;
char *name;
static struct mainloop_fd_callbacks lrmd_tls_callbacks = {
.dispatch = lrmd_tls_dispatch,
.destroy = lrmd_tls_connection_destroy,
};
gnutls_datum_t psk_key = { NULL, 0 };
native->async_timer = 0;
if (rc != pcmk_rc_ok) {
lrmd_tls_connection_destroy(lrmd);
crm_info("Could not connect to Pacemaker Remote at %s:%d: %s "
CRM_XS " rc=%d",
native->server, native->port, pcmk_rc_str(rc), rc);
report_async_connection_result(lrmd, pcmk_rc2legacy(rc));
return;
}
/* The TCP connection was successful, so establish the TLS connection.
* @TODO make this async to avoid blocking code in client
*/
native->sock = sock;
rc = lrmd_tls_set_key(&psk_key);
if (rc != 0) {
crm_warn("Could not set key for Pacemaker Remote at %s:%d " CRM_XS " rc=%d",
native->server, native->port, rc);
lrmd_tls_connection_destroy(lrmd);
report_async_connection_result(lrmd, rc);
return;
}
gnutls_psk_allocate_client_credentials(&native->psk_cred_c);
gnutls_psk_set_client_credentials(native->psk_cred_c, DEFAULT_REMOTE_USERNAME, &psk_key, GNUTLS_PSK_KEY_RAW);
gnutls_free(psk_key.data);
native->remote->tls_session = pcmk__new_tls_session(sock, GNUTLS_CLIENT,
GNUTLS_CRD_PSK,
native->psk_cred_c);
if (native->remote->tls_session == NULL) {
lrmd_tls_connection_destroy(lrmd);
report_async_connection_result(lrmd, -EPROTO);
return;
}
if (lrmd__tls_client_handshake(native->remote) != pcmk_rc_ok) {
crm_warn("Disconnecting after TLS handshake with Pacemaker Remote server %s:%d failed",
native->server, native->port);
gnutls_deinit(*native->remote->tls_session);
gnutls_free(native->remote->tls_session);
native->remote->tls_session = NULL;
lrmd_tls_connection_destroy(lrmd);
report_async_connection_result(lrmd, -EKEYREJECTED);
return;
}
crm_info("TLS connection to Pacemaker Remote server %s:%d succeeded",
native->server, native->port);
name = crm_strdup_printf("pacemaker-remote-%s:%d",
native->server, native->port);
native->process_notify = mainloop_add_trigger(G_PRIORITY_HIGH, lrmd_tls_dispatch, lrmd);
native->source =
mainloop_add_fd(name, G_PRIORITY_HIGH, native->sock, lrmd, &lrmd_tls_callbacks);
rc = lrmd_handshake(lrmd, name);
free(name);
report_async_connection_result(lrmd, rc);
return;
}
static int
lrmd_tls_connect_async(lrmd_t * lrmd, int timeout /*ms */ )
{
int rc;
int timer_id = 0;
lrmd_private_t *native = lrmd->lrmd_private;
lrmd_gnutls_global_init();
native->sock = -1;
rc = pcmk__connect_remote(native->server, native->port, timeout, &timer_id,
&(native->sock), lrmd, lrmd_tcp_connect_cb);
if (rc != pcmk_rc_ok) {
crm_warn("Pacemaker Remote connection to %s:%s failed: %s "
CRM_XS " rc=%d",
native->server, native->port, pcmk_rc_str(rc), rc);
return -1;
}
native->async_timer = timer_id;
return pcmk_ok;
}
static int
lrmd_tls_connect(lrmd_t * lrmd, int *fd)
{
static struct mainloop_fd_callbacks lrmd_tls_callbacks = {
.dispatch = lrmd_tls_dispatch,
.destroy = lrmd_tls_connection_destroy,
};
int rc;
lrmd_private_t *native = lrmd->lrmd_private;
gnutls_datum_t psk_key = { NULL, 0 };
lrmd_gnutls_global_init();
native->sock = -1;
rc = pcmk__connect_remote(native->server, native->port, 0, NULL,
&(native->sock), NULL, NULL);
if (rc != pcmk_rc_ok) {
crm_warn("Pacemaker Remote connection to %s:%s failed: %s "
CRM_XS " rc=%d",
native->server, native->port, pcmk_rc_str(rc), rc);
lrmd_tls_connection_destroy(lrmd);
return -ENOTCONN;
}
rc = lrmd_tls_set_key(&psk_key);
if (rc < 0) {
lrmd_tls_connection_destroy(lrmd);
return rc;
}
gnutls_psk_allocate_client_credentials(&native->psk_cred_c);
gnutls_psk_set_client_credentials(native->psk_cred_c, DEFAULT_REMOTE_USERNAME, &psk_key, GNUTLS_PSK_KEY_RAW);
gnutls_free(psk_key.data);
native->remote->tls_session = pcmk__new_tls_session(native->sock, GNUTLS_CLIENT,
GNUTLS_CRD_PSK,
native->psk_cred_c);
if (native->remote->tls_session == NULL) {
lrmd_tls_connection_destroy(lrmd);
return -EPROTO;
}
if (lrmd__tls_client_handshake(native->remote) != pcmk_rc_ok) {
crm_err("Session creation for %s:%d failed", native->server, native->port);
gnutls_deinit(*native->remote->tls_session);
gnutls_free(native->remote->tls_session);
native->remote->tls_session = NULL;
lrmd_tls_connection_destroy(lrmd);
return -EKEYREJECTED;
}
crm_info("Client TLS connection established with Pacemaker Remote server %s:%d", native->server,
native->port);
if (fd) {
*fd = native->sock;
} else {
char *name = crm_strdup_printf("pacemaker-remote-%s:%d",
native->server, native->port);
native->process_notify = mainloop_add_trigger(G_PRIORITY_HIGH, lrmd_tls_dispatch, lrmd);
native->source =
mainloop_add_fd(name, G_PRIORITY_HIGH, native->sock, lrmd, &lrmd_tls_callbacks);
free(name);
}
return pcmk_ok;
}
#endif
static int
lrmd_api_connect(lrmd_t * lrmd, const char *name, int *fd)
{
int rc = -ENOTCONN;
lrmd_private_t *native = lrmd->lrmd_private;
switch (native->type) {
case PCMK__CLIENT_IPC:
rc = lrmd_ipc_connect(lrmd, fd);
break;
#ifdef HAVE_GNUTLS_GNUTLS_H
case PCMK__CLIENT_TLS:
rc = lrmd_tls_connect(lrmd, fd);
break;
#endif
default:
crm_err("Unsupported connection type: %d", native->type);
}
if (rc == pcmk_ok) {
rc = lrmd_handshake(lrmd, name);
}
return rc;
}
static int
lrmd_api_connect_async(lrmd_t * lrmd, const char *name, int timeout)
{
int rc = 0;
lrmd_private_t *native = lrmd->lrmd_private;
CRM_CHECK(native && native->callback, return -1);
switch (native->type) {
case PCMK__CLIENT_IPC:
/* fake async connection with ipc. it should be fast
* enough that we gain very little from async */
rc = lrmd_api_connect(lrmd, name, NULL);
if (!rc) {
report_async_connection_result(lrmd, rc);
}
break;
#ifdef HAVE_GNUTLS_GNUTLS_H
case PCMK__CLIENT_TLS:
rc = lrmd_tls_connect_async(lrmd, timeout);
if (rc) {
/* connection failed, report rc now */
report_async_connection_result(lrmd, rc);
}
break;
#endif
default:
crm_err("Unsupported connection type: %d", native->type);
}
return rc;
}
static void
lrmd_ipc_disconnect(lrmd_t * lrmd)
{
lrmd_private_t *native = lrmd->lrmd_private;
if (native->source != NULL) {
/* Attached to mainloop */
mainloop_del_ipc_client(native->source);
native->source = NULL;
native->ipc = NULL;
} else if (native->ipc) {
/* Not attached to mainloop */
crm_ipc_t *ipc = native->ipc;
native->ipc = NULL;
crm_ipc_close(ipc);
crm_ipc_destroy(ipc);
}
}
#ifdef HAVE_GNUTLS_GNUTLS_H
static void
lrmd_tls_disconnect(lrmd_t * lrmd)
{
lrmd_private_t *native = lrmd->lrmd_private;
if (native->remote->tls_session) {
gnutls_bye(*native->remote->tls_session, GNUTLS_SHUT_RDWR);
gnutls_deinit(*native->remote->tls_session);
gnutls_free(native->remote->tls_session);
native->remote->tls_session = 0;
}
if (native->async_timer) {
g_source_remove(native->async_timer);
native->async_timer = 0;
}
if (native->source != NULL) {
/* Attached to mainloop */
mainloop_del_ipc_client(native->source);
native->source = NULL;
} else if (native->sock) {
close(native->sock);
native->sock = 0;
}
if (native->pending_notify) {
g_list_free_full(native->pending_notify, lrmd_free_xml);
native->pending_notify = NULL;
}
}
#endif
static int
lrmd_api_disconnect(lrmd_t * lrmd)
{
lrmd_private_t *native = lrmd->lrmd_private;
crm_info("Disconnecting %s %s executor connection",
pcmk__client_type_str(native->type),
(native->remote_nodename? native->remote_nodename : "local"));
switch (native->type) {
case PCMK__CLIENT_IPC:
lrmd_ipc_disconnect(lrmd);
break;
#ifdef HAVE_GNUTLS_GNUTLS_H
case PCMK__CLIENT_TLS:
lrmd_tls_disconnect(lrmd);
break;
#endif
default:
crm_err("Unsupported connection type: %d", native->type);
}
free(native->token);
native->token = NULL;
free(native->peer_version);
native->peer_version = NULL;
return 0;
}
static int
lrmd_api_register_rsc(lrmd_t * lrmd,
const char *rsc_id,
const char *class,
const char *provider, const char *type, enum lrmd_call_options options)
{
int rc = pcmk_ok;
xmlNode *data = NULL;
if (!class || !type || !rsc_id) {
return -EINVAL;
}
if (is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider) && !provider) {
return -EINVAL;
}
data = create_xml_node(NULL, F_LRMD_RSC);
crm_xml_add(data, F_LRMD_ORIGIN, __FUNCTION__);
crm_xml_add(data, F_LRMD_RSC_ID, rsc_id);
crm_xml_add(data, F_LRMD_CLASS, class);
crm_xml_add(data, F_LRMD_PROVIDER, provider);
crm_xml_add(data, F_LRMD_TYPE, type);
rc = lrmd_send_command(lrmd, LRMD_OP_RSC_REG, data, NULL, 0, options, TRUE);
free_xml(data);
return rc;
}
static int
lrmd_api_unregister_rsc(lrmd_t * lrmd, const char *rsc_id, enum lrmd_call_options options)
{
int rc = pcmk_ok;
xmlNode *data = create_xml_node(NULL, F_LRMD_RSC);
crm_xml_add(data, F_LRMD_ORIGIN, __FUNCTION__);
crm_xml_add(data, F_LRMD_RSC_ID, rsc_id);
rc = lrmd_send_command(lrmd, LRMD_OP_RSC_UNREG, data, NULL, 0, options, TRUE);
free_xml(data);
return rc;
}
lrmd_rsc_info_t *
lrmd_new_rsc_info(const char *rsc_id, const char *standard,
const char *provider, const char *type)
{
lrmd_rsc_info_t *rsc_info = calloc(1, sizeof(lrmd_rsc_info_t));
CRM_ASSERT(rsc_info);
if (rsc_id) {
rsc_info->id = strdup(rsc_id);
CRM_ASSERT(rsc_info->id);
}
if (standard) {
rsc_info->standard = strdup(standard);
CRM_ASSERT(rsc_info->standard);
}
if (provider) {
rsc_info->provider = strdup(provider);
CRM_ASSERT(rsc_info->provider);
}
if (type) {
rsc_info->type = strdup(type);
CRM_ASSERT(rsc_info->type);
}
return rsc_info;
}
lrmd_rsc_info_t *
lrmd_copy_rsc_info(lrmd_rsc_info_t * rsc_info)
{
return lrmd_new_rsc_info(rsc_info->id, rsc_info->standard,
rsc_info->provider, rsc_info->type);
}
void
lrmd_free_rsc_info(lrmd_rsc_info_t * rsc_info)
{
if (!rsc_info) {
return;
}
free(rsc_info->id);
free(rsc_info->type);
free(rsc_info->standard);
free(rsc_info->provider);
free(rsc_info);
}
static lrmd_rsc_info_t *
lrmd_api_get_rsc_info(lrmd_t * lrmd, const char *rsc_id, enum lrmd_call_options options)
{
lrmd_rsc_info_t *rsc_info = NULL;
xmlNode *data = create_xml_node(NULL, F_LRMD_RSC);
xmlNode *output = NULL;
const char *class = NULL;
const char *provider = NULL;
const char *type = NULL;
crm_xml_add(data, F_LRMD_ORIGIN, __FUNCTION__);
crm_xml_add(data, F_LRMD_RSC_ID, rsc_id);
lrmd_send_command(lrmd, LRMD_OP_RSC_INFO, data, &output, 0, options, TRUE);
free_xml(data);
if (!output) {
return NULL;
}
class = crm_element_value(output, F_LRMD_CLASS);
provider = crm_element_value(output, F_LRMD_PROVIDER);
type = crm_element_value(output, F_LRMD_TYPE);
if (!class || !type) {
free_xml(output);
return NULL;
} else if (is_set(pcmk_get_ra_caps(class), pcmk_ra_cap_provider)
&& !provider) {
free_xml(output);
return NULL;
}
rsc_info = lrmd_new_rsc_info(rsc_id, class, provider, type);
free_xml(output);
return rsc_info;
}
void
lrmd_free_op_info(lrmd_op_info_t *op_info)
{
if (op_info) {
free(op_info->rsc_id);
free(op_info->action);
free(op_info->interval_ms_s);
free(op_info->timeout_ms_s);
free(op_info);
}
}
static int
lrmd_api_get_recurring_ops(lrmd_t *lrmd, const char *rsc_id, int timeout_ms,
enum lrmd_call_options options, GList **output)
{
xmlNode *data = NULL;
xmlNode *output_xml = NULL;
int rc = pcmk_ok;
if (output == NULL) {
return -EINVAL;
}
*output = NULL;
// Send request
if (rsc_id) {
data = create_xml_node(NULL, F_LRMD_RSC);
crm_xml_add(data, F_LRMD_ORIGIN, __FUNCTION__);
crm_xml_add(data, F_LRMD_RSC_ID, rsc_id);
}
rc = lrmd_send_command(lrmd, LRMD_OP_GET_RECURRING, data, &output_xml,
timeout_ms, options, TRUE);
if (data) {
free_xml(data);
}
// Process reply
if ((rc != pcmk_ok) || (output_xml == NULL)) {
return rc;
}
for (xmlNode *rsc_xml = first_named_child(output_xml, F_LRMD_RSC);
rsc_xml != NULL; rsc_xml = crm_next_same_xml(rsc_xml)) {
rsc_id = crm_element_value(rsc_xml, F_LRMD_RSC_ID);
if (rsc_id == NULL) {
crm_err("Could not parse recurring operation information from executor");
continue;
}
for (xmlNode *op_xml = first_named_child(rsc_xml, T_LRMD_RSC_OP);
op_xml != NULL; op_xml = crm_next_same_xml(op_xml)) {
lrmd_op_info_t *op_info = calloc(1, sizeof(lrmd_op_info_t));
CRM_CHECK(op_info != NULL, break);
op_info->rsc_id = strdup(rsc_id);
op_info->action = crm_element_value_copy(op_xml, F_LRMD_RSC_ACTION);
op_info->interval_ms_s = crm_element_value_copy(op_xml,
F_LRMD_RSC_INTERVAL);
op_info->timeout_ms_s = crm_element_value_copy(op_xml,
F_LRMD_TIMEOUT);
*output = g_list_prepend(*output, op_info);
}
}
free_xml(output_xml);
return rc;
}
static void
lrmd_api_set_callback(lrmd_t * lrmd, lrmd_event_callback callback)
{
lrmd_private_t *native = lrmd->lrmd_private;
native->callback = callback;
}
void
lrmd_internal_set_proxy_callback(lrmd_t * lrmd, void *userdata, void (*callback)(lrmd_t *lrmd, void *userdata, xmlNode *msg))
{
lrmd_private_t *native = lrmd->lrmd_private;
native->proxy_callback = callback;
native->proxy_callback_userdata = userdata;
}
void
lrmd_internal_proxy_dispatch(lrmd_t *lrmd, xmlNode *msg)
{
lrmd_private_t *native = lrmd->lrmd_private;
if (native->proxy_callback) {
crm_log_xml_trace(msg, "PROXY_INBOUND");
native->proxy_callback(lrmd, native->proxy_callback_userdata, msg);
}
}
int
lrmd_internal_proxy_send(lrmd_t * lrmd, xmlNode *msg)
{
if (lrmd == NULL) {
return -ENOTCONN;
}
crm_xml_add(msg, F_LRMD_OPERATION, CRM_OP_IPC_FWD);
crm_log_xml_trace(msg, "PROXY_OUTBOUND");
return lrmd_send_xml_no_reply(lrmd, msg);
}
static int
stonith_get_metadata(const char *provider, const char *type, char **output)
{
int rc = pcmk_ok;
stonith_t *stonith_api = stonith_api_new();
if (stonith_api == NULL) {
crm_err("Could not get fence agent meta-data: API memory allocation failed");
return -ENOMEM;
}
rc = stonith_api->cmds->metadata(stonith_api, st_opt_sync_call, type,
provider, output, 0);
if ((rc == pcmk_ok) && (*output == NULL)) {
rc = -EIO;
}
stonith_api->cmds->free(stonith_api);
return rc;
}
static int
lrmd_api_get_metadata(lrmd_t *lrmd, const char *standard, const char *provider,
const char *type, char **output,
enum lrmd_call_options options)
{
return lrmd->cmds->get_metadata_params(lrmd, standard, provider, type,
output, options, NULL);
}
static int
lrmd_api_get_metadata_params(lrmd_t *lrmd, const char *standard,
const char *provider, const char *type,
char **output, enum lrmd_call_options options,
lrmd_key_value_t *params)
{
svc_action_t *action = NULL;
GHashTable *params_table = NULL;
if (!standard || !type) {
lrmd_key_value_freeall(params);
return -EINVAL;
}
if (safe_str_eq(standard, PCMK_RESOURCE_CLASS_STONITH)) {
lrmd_key_value_freeall(params);
return stonith_get_metadata(provider, type, output);
}
params_table = crm_str_table_new();
for (const lrmd_key_value_t *param = params; param; param = param->next) {
g_hash_table_insert(params_table, strdup(param->key), strdup(param->value));
}
action = resources_action_create(type, standard, provider, type,
CRMD_ACTION_METADATA, 0,
CRMD_METADATA_CALL_TIMEOUT, params_table,
0);
lrmd_key_value_freeall(params);
if (action == NULL) {
crm_err("Unable to retrieve meta-data for %s:%s:%s",
standard, provider, type);
return -EINVAL;
}
if (!services_action_sync(action)) {
crm_err("Failed to retrieve meta-data for %s:%s:%s",
standard, provider, type);
services_action_free(action);
return -EIO;
}
if (!action->stdout_data) {
crm_err("Failed to receive meta-data for %s:%s:%s",
standard, provider, type);
services_action_free(action);
return -EIO;
}
*output = strdup(action->stdout_data);
services_action_free(action);
return pcmk_ok;
}
static int
lrmd_api_exec(lrmd_t *lrmd, const char *rsc_id, const char *action,
const char *userdata, guint interval_ms,
int timeout, /* ms */
int start_delay, /* ms */
enum lrmd_call_options options, lrmd_key_value_t * params)
{
int rc = pcmk_ok;
xmlNode *data = create_xml_node(NULL, F_LRMD_RSC);
xmlNode *args = create_xml_node(data, XML_TAG_ATTRS);
lrmd_key_value_t *tmp = NULL;
crm_xml_add(data, F_LRMD_ORIGIN, __FUNCTION__);
crm_xml_add(data, F_LRMD_RSC_ID, rsc_id);
crm_xml_add(data, F_LRMD_RSC_ACTION, action);
crm_xml_add(data, F_LRMD_RSC_USERDATA_STR, userdata);
crm_xml_add_ms(data, F_LRMD_RSC_INTERVAL, interval_ms);
crm_xml_add_int(data, F_LRMD_TIMEOUT, timeout);
crm_xml_add_int(data, F_LRMD_RSC_START_DELAY, start_delay);
for (tmp = params; tmp; tmp = tmp->next) {
hash2smartfield((gpointer) tmp->key, (gpointer) tmp->value, args);
}
rc = lrmd_send_command(lrmd, LRMD_OP_RSC_EXEC, data, NULL, timeout, options, TRUE);
free_xml(data);
lrmd_key_value_freeall(params);
return rc;
}
/* timeout is in ms */
static int
lrmd_api_exec_alert(lrmd_t *lrmd, const char *alert_id, const char *alert_path,
int timeout, lrmd_key_value_t *params)
{
int rc = pcmk_ok;
xmlNode *data = create_xml_node(NULL, F_LRMD_ALERT);
xmlNode *args = create_xml_node(data, XML_TAG_ATTRS);
lrmd_key_value_t *tmp = NULL;
crm_xml_add(data, F_LRMD_ORIGIN, __FUNCTION__);
crm_xml_add(data, F_LRMD_ALERT_ID, alert_id);
crm_xml_add(data, F_LRMD_ALERT_PATH, alert_path);
crm_xml_add_int(data, F_LRMD_TIMEOUT, timeout);
for (tmp = params; tmp; tmp = tmp->next) {
hash2smartfield((gpointer) tmp->key, (gpointer) tmp->value, args);
}
rc = lrmd_send_command(lrmd, LRMD_OP_ALERT_EXEC, data, NULL, timeout,
lrmd_opt_notify_orig_only, TRUE);
free_xml(data);
lrmd_key_value_freeall(params);
return rc;
}
static int
lrmd_api_cancel(lrmd_t *lrmd, const char *rsc_id, const char *action,
guint interval_ms)
{
int rc = pcmk_ok;
xmlNode *data = create_xml_node(NULL, F_LRMD_RSC);
crm_xml_add(data, F_LRMD_ORIGIN, __FUNCTION__);
crm_xml_add(data, F_LRMD_RSC_ACTION, action);
crm_xml_add(data, F_LRMD_RSC_ID, rsc_id);
crm_xml_add_ms(data, F_LRMD_RSC_INTERVAL, interval_ms);
rc = lrmd_send_command(lrmd, LRMD_OP_RSC_CANCEL, data, NULL, 0, 0, TRUE);
free_xml(data);
return rc;
}
static int
list_stonith_agents(lrmd_list_t ** resources)
{
int rc = 0;
stonith_t *stonith_api = stonith_api_new();
stonith_key_value_t *stonith_resources = NULL;
stonith_key_value_t *dIter = NULL;
if (stonith_api == NULL) {
crm_err("Could not list fence agents: API memory allocation failed");
return -ENOMEM;
}
stonith_api->cmds->list_agents(stonith_api, st_opt_sync_call, NULL,
&stonith_resources, 0);
stonith_api->cmds->free(stonith_api);
for (dIter = stonith_resources; dIter; dIter = dIter->next) {
rc++;
if (resources) {
*resources = lrmd_list_add(*resources, dIter->value);
}
}
stonith_key_value_freeall(stonith_resources, 1, 0);
return rc;
}
static int
lrmd_api_list_agents(lrmd_t * lrmd, lrmd_list_t ** resources, const char *class,
const char *provider)
{
int rc = 0;
int stonith_count = 0; // Initially, whether to include stonith devices
if (safe_str_eq(class, PCMK_RESOURCE_CLASS_STONITH)) {
stonith_count = 1;
} else {
GListPtr gIter = NULL;
GList *agents = resources_list_agents(class, provider);
for (gIter = agents; gIter != NULL; gIter = gIter->next) {
*resources = lrmd_list_add(*resources, (const char *)gIter->data);
rc++;
}
g_list_free_full(agents, free);
if (!class) {
stonith_count = 1;
}
}
if (stonith_count) {
// Now, if stonith devices are included, how many there are
stonith_count = list_stonith_agents(resources);
if (stonith_count > 0) {
rc += stonith_count;
}
}
if (rc == 0) {
crm_notice("No agents found for class %s", class);
rc = -EPROTONOSUPPORT;
}
return rc;
}
static int
does_provider_have_agent(const char *agent, const char *provider, const char *class)
{
int found = 0;
GList *agents = NULL;
GListPtr gIter2 = NULL;
agents = resources_list_agents(class, provider);
for (gIter2 = agents; gIter2 != NULL; gIter2 = gIter2->next) {
if (safe_str_eq(agent, gIter2->data)) {
found = 1;
}
}
g_list_free_full(agents, free);
return found;
}
static int
lrmd_api_list_ocf_providers(lrmd_t * lrmd, const char *agent, lrmd_list_t ** providers)
{
int rc = pcmk_ok;
char *provider = NULL;
GList *ocf_providers = NULL;
GListPtr gIter = NULL;
ocf_providers = resources_list_providers(PCMK_RESOURCE_CLASS_OCF);
for (gIter = ocf_providers; gIter != NULL; gIter = gIter->next) {
provider = gIter->data;
if (!agent || does_provider_have_agent(agent, provider,
PCMK_RESOURCE_CLASS_OCF)) {
*providers = lrmd_list_add(*providers, (const char *)gIter->data);
rc++;
}
}
g_list_free_full(ocf_providers, free);
return rc;
}
static int
lrmd_api_list_standards(lrmd_t * lrmd, lrmd_list_t ** supported)
{
int rc = 0;
GList *standards = NULL;
GListPtr gIter = NULL;
standards = resources_list_standards();
for (gIter = standards; gIter != NULL; gIter = gIter->next) {
*supported = lrmd_list_add(*supported, (const char *)gIter->data);
rc++;
}
if (list_stonith_agents(NULL) > 0) {
*supported = lrmd_list_add(*supported, PCMK_RESOURCE_CLASS_STONITH);
rc++;
}
g_list_free_full(standards, free);
return rc;
}
lrmd_t *
lrmd_api_new(void)
{
lrmd_t *new_lrmd = NULL;
lrmd_private_t *pvt = NULL;
new_lrmd = calloc(1, sizeof(lrmd_t));
pvt = calloc(1, sizeof(lrmd_private_t));
pvt->remote = calloc(1, sizeof(pcmk__remote_t));
new_lrmd->cmds = calloc(1, sizeof(lrmd_api_operations_t));
pvt->type = PCMK__CLIENT_IPC;
new_lrmd->lrmd_private = pvt;
new_lrmd->cmds->connect = lrmd_api_connect;
new_lrmd->cmds->connect_async = lrmd_api_connect_async;
new_lrmd->cmds->is_connected = lrmd_api_is_connected;
new_lrmd->cmds->poke_connection = lrmd_api_poke_connection;
new_lrmd->cmds->disconnect = lrmd_api_disconnect;
new_lrmd->cmds->register_rsc = lrmd_api_register_rsc;
new_lrmd->cmds->unregister_rsc = lrmd_api_unregister_rsc;
new_lrmd->cmds->get_rsc_info = lrmd_api_get_rsc_info;
new_lrmd->cmds->get_recurring_ops = lrmd_api_get_recurring_ops;
new_lrmd->cmds->set_callback = lrmd_api_set_callback;
new_lrmd->cmds->get_metadata = lrmd_api_get_metadata;
new_lrmd->cmds->exec = lrmd_api_exec;
new_lrmd->cmds->cancel = lrmd_api_cancel;
new_lrmd->cmds->list_agents = lrmd_api_list_agents;
new_lrmd->cmds->list_ocf_providers = lrmd_api_list_ocf_providers;
new_lrmd->cmds->list_standards = lrmd_api_list_standards;
new_lrmd->cmds->exec_alert = lrmd_api_exec_alert;
new_lrmd->cmds->get_metadata_params = lrmd_api_get_metadata_params;
return new_lrmd;
}
lrmd_t *
lrmd_remote_api_new(const char *nodename, const char *server, int port)
{
#ifdef HAVE_GNUTLS_GNUTLS_H
lrmd_t *new_lrmd = lrmd_api_new();
lrmd_private_t *native = new_lrmd->lrmd_private;
if (!nodename && !server) {
lrmd_api_delete(new_lrmd);
return NULL;
}
native->type = PCMK__CLIENT_TLS;
native->remote_nodename = nodename ? strdup(nodename) : strdup(server);
native->server = server ? strdup(server) : strdup(nodename);
native->port = port;
if (native->port == 0) {
native->port = crm_default_remote_port();
}
return new_lrmd;
#else
crm_err("Cannot communicate with Pacemaker Remote because GnuTLS is not enabled for this build");
return NULL;
#endif
}
void
lrmd_api_delete(lrmd_t * lrmd)
{
if (!lrmd) {
return;
}
lrmd->cmds->disconnect(lrmd); /* no-op if already disconnected */
free(lrmd->cmds);
if (lrmd->lrmd_private) {
lrmd_private_t *native = lrmd->lrmd_private;
#ifdef HAVE_GNUTLS_GNUTLS_H
free(native->server);
#endif
free(native->remote_nodename);
free(native->remote);
free(native->token);
free(native->peer_version);
}
free(lrmd->lrmd_private);
free(lrmd);
}
diff --git a/lib/pengine/utils.c b/lib/pengine/utils.c
index 584def405b..a82acea5e0 100644
--- a/lib/pengine/utils.c
+++ b/lib/pengine/utils.c
@@ -1,2808 +1,2807 @@
/*
* Copyright 2004-2020 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.
*/
#include <crm_internal.h>
#include <crm/crm.h>
#include <crm/msg_xml.h>
#include <crm/common/xml.h>
#include <crm/common/util.h>
#include <glib.h>
#include <stdbool.h>
#include <crm/pengine/rules.h>
#include <crm/pengine/internal.h>
extern xmlNode *get_object_root(const char *object_type, xmlNode * the_root);
void print_str_str(gpointer key, gpointer value, gpointer user_data);
gboolean ghash_free_str_str(gpointer key, gpointer value, gpointer user_data);
static void unpack_operation(pe_action_t * action, xmlNode * xml_obj, pe_resource_t * container,
pe_working_set_t * data_set, guint interval_ms);
static xmlNode *find_rsc_op_entry_helper(pe_resource_t * rsc, const char *key,
gboolean include_disabled);
#if ENABLE_VERSIONED_ATTRS
pe_rsc_action_details_t *
pe_rsc_action_details(pe_action_t *action)
{
pe_rsc_action_details_t *details;
CRM_CHECK(action != NULL, return NULL);
if (action->action_details == NULL) {
action->action_details = calloc(1, sizeof(pe_rsc_action_details_t));
CRM_CHECK(action->action_details != NULL, return NULL);
}
details = (pe_rsc_action_details_t *) action->action_details;
if (details->versioned_parameters == NULL) {
details->versioned_parameters = create_xml_node(NULL,
XML_TAG_OP_VER_ATTRS);
}
if (details->versioned_meta == NULL) {
details->versioned_meta = create_xml_node(NULL, XML_TAG_OP_VER_META);
}
return details;
}
static void
pe_free_rsc_action_details(pe_action_t *action)
{
pe_rsc_action_details_t *details;
if ((action == NULL) || (action->action_details == NULL)) {
return;
}
details = (pe_rsc_action_details_t *) action->action_details;
if (details->versioned_parameters) {
free_xml(details->versioned_parameters);
}
if (details->versioned_meta) {
free_xml(details->versioned_meta);
}
action->action_details = NULL;
}
#endif
/*!
* \internal
* \brief Check whether we can fence a particular node
*
* \param[in] data_set Working set for cluster
* \param[in] node Name of node to check
*
* \return true if node can be fenced, false otherwise
*/
bool
pe_can_fence(pe_working_set_t *data_set, pe_node_t *node)
{
if (pe__is_guest_node(node)) {
/* Guest nodes are fenced by stopping their container resource. We can
* do that if the container's host is either online or fenceable.
*/
pe_resource_t *rsc = node->details->remote_rsc->container;
for (GList *n = rsc->running_on; n != NULL; n = n->next) {
pe_node_t *container_node = n->data;
if (!container_node->details->online
&& !pe_can_fence(data_set, container_node)) {
return false;
}
}
return true;
} else if(is_not_set(data_set->flags, pe_flag_stonith_enabled)) {
return false; /* Turned off */
} else if (is_not_set(data_set->flags, pe_flag_have_stonith_resource)) {
return false; /* No devices */
} else if (is_set(data_set->flags, pe_flag_have_quorum)) {
return true;
} else if (data_set->no_quorum_policy == no_quorum_ignore) {
return true;
} else if(node == NULL) {
return false;
} else if(node->details->online) {
crm_notice("We can fence %s without quorum because they're in our membership", node->details->uname);
return true;
}
crm_trace("Cannot fence %s", node->details->uname);
return false;
}
/*!
* \internal
* \brief Copy a node object
*
* \param[in] this_node Node object to copy
*
* \return Newly allocated shallow copy of this_node
* \note This function asserts on errors and is guaranteed to return non-NULL.
*/
pe_node_t *
pe__copy_node(const pe_node_t *this_node)
{
pe_node_t *new_node = NULL;
CRM_ASSERT(this_node != NULL);
new_node = calloc(1, sizeof(pe_node_t));
CRM_ASSERT(new_node != NULL);
new_node->rsc_discover_mode = this_node->rsc_discover_mode;
new_node->weight = this_node->weight;
new_node->fixed = this_node->fixed;
new_node->details = this_node->details;
return new_node;
}
/* any node in list1 or list2 and not in the other gets a score of -INFINITY */
void
node_list_exclude(GHashTable * hash, GListPtr list, gboolean merge_scores)
{
GHashTable *result = hash;
pe_node_t *other_node = NULL;
GListPtr gIter = list;
GHashTableIter iter;
pe_node_t *node = NULL;
g_hash_table_iter_init(&iter, hash);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
other_node = pe_find_node_id(list, node->details->id);
if (other_node == NULL) {
node->weight = -INFINITY;
} else if (merge_scores) {
node->weight = pe__add_scores(node->weight, other_node->weight);
}
}
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
other_node = pe_hash_table_lookup(result, node->details->id);
if (other_node == NULL) {
pe_node_t *new_node = pe__copy_node(node);
new_node->weight = -INFINITY;
g_hash_table_insert(result, (gpointer) new_node->details->id, new_node);
}
}
}
/*!
* \internal
* \brief Create a node hash table from a node list
*
* \param[in] list Node list
*
* \return Hash table equivalent of node list
*/
GHashTable *
pe__node_list2table(GList *list)
{
GHashTable *result = NULL;
result = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free);
for (GList *gIter = list; gIter != NULL; gIter = gIter->next) {
pe_node_t *new_node = pe__copy_node((pe_node_t *) gIter->data);
g_hash_table_insert(result, (gpointer) new_node->details->id, new_node);
}
return result;
}
gint
sort_node_uname(gconstpointer a, gconstpointer b)
{
return pcmk_numeric_strcasecmp(((const pe_node_t *) a)->details->uname,
((const pe_node_t *) b)->details->uname);
}
/*!
* \internal
* \brief Output node weights to stdout
*
* \param[in] rsc Use allowed nodes for this resource
* \param[in] comment Text description to prefix lines with
* \param[in] nodes If rsc is not specified, use these nodes
*/
static void
pe__output_node_weights(pe_resource_t *rsc, const char *comment,
GHashTable *nodes)
{
char score[128]; // Stack-allocated since this is called frequently
// Sort the nodes so the output is consistent for regression tests
GList *list = g_list_sort(g_hash_table_get_values(nodes), sort_node_uname);
for (GList *gIter = list; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
score2char_stack(node->weight, score, sizeof(score));
if (rsc) {
printf("%s: %s allocation score on %s: %s\n",
comment, rsc->id, node->details->uname, score);
} else {
printf("%s: %s = %s\n", comment, node->details->uname, score);
}
}
g_list_free(list);
}
/*!
* \internal
* \brief Log node weights at trace level
*
* \param[in] file Caller's filename
* \param[in] function Caller's function name
* \param[in] line Caller's line number
* \param[in] rsc Use allowed nodes for this resource
* \param[in] comment Text description to prefix lines with
* \param[in] nodes If rsc is not specified, use these nodes
*/
static void
pe__log_node_weights(const char *file, const char *function, int line,
pe_resource_t *rsc, const char *comment, GHashTable *nodes)
{
GHashTableIter iter;
pe_node_t *node = NULL;
char score[128]; // Stack-allocated since this is called frequently
// Don't waste time if we're not tracing at this point
pcmk__log_else(LOG_TRACE, return);
g_hash_table_iter_init(&iter, nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **) &node)) {
score2char_stack(node->weight, score, sizeof(score));
if (rsc) {
qb_log_from_external_source(function, file,
"%s: %s allocation score on %s: %s",
LOG_TRACE, line, 0,
comment, rsc->id,
node->details->uname, score);
} else {
qb_log_from_external_source(function, file, "%s: %s = %s",
LOG_TRACE, line, 0,
comment, node->details->uname,
score);
}
}
}
/*!
* \internal
* \brief Log or output node weights
*
* \param[in] file Caller's filename
* \param[in] function Caller's function name
* \param[in] line Caller's line number
* \param[in] to_log Log if true, otherwise output
* \param[in] rsc Use allowed nodes for this resource
* \param[in] comment Text description to prefix lines with
* \param[in] nodes If rsc is not specified, use these nodes
*/
void
pe__show_node_weights_as(const char *file, const char *function, int line,
bool to_log, pe_resource_t *rsc, const char *comment,
GHashTable *nodes)
{
if (rsc != NULL) {
if (is_set(rsc->flags, pe_rsc_orphan)) {
// Don't show allocation scores for orphans
return;
}
nodes = rsc->allowed_nodes;
}
if (nodes == NULL) {
// Nothing to show
return;
}
if (to_log) {
pe__log_node_weights(file, function, line, rsc, comment, nodes);
} else {
pe__output_node_weights(rsc, comment, nodes);
}
// If this resource has children, repeat recursively for each
if (rsc && rsc->children) {
for (GList *gIter = rsc->children; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child = (pe_resource_t *) gIter->data;
pe__show_node_weights_as(file, function, line, to_log, child,
comment, nodes);
}
}
}
static void
append_dump_text(gpointer key, gpointer value, gpointer user_data)
{
char **dump_text = user_data;
char *new_text = crm_strdup_printf("%s %s=%s",
*dump_text, (char *)key, (char *)value);
free(*dump_text);
*dump_text = new_text;
}
void
dump_node_capacity(int level, const char *comment, pe_node_t * node)
{
char *dump_text = crm_strdup_printf("%s: %s capacity:",
comment, node->details->uname);
g_hash_table_foreach(node->details->utilization, append_dump_text, &dump_text);
if (level == LOG_STDOUT) {
fprintf(stdout, "%s\n", dump_text);
} else {
crm_trace("%s", dump_text);
}
free(dump_text);
}
void
dump_rsc_utilization(int level, const char *comment, pe_resource_t * rsc, pe_node_t * node)
{
char *dump_text = crm_strdup_printf("%s: %s utilization on %s:",
comment, rsc->id, node->details->uname);
g_hash_table_foreach(rsc->utilization, append_dump_text, &dump_text);
switch (level) {
case LOG_STDOUT:
fprintf(stdout, "%s\n", dump_text);
break;
case LOG_NEVER:
break;
default:
crm_trace("%s", dump_text);
}
free(dump_text);
}
gint
sort_rsc_index(gconstpointer a, gconstpointer b)
{
const pe_resource_t *resource1 = (const pe_resource_t *)a;
const pe_resource_t *resource2 = (const pe_resource_t *)b;
if (a == NULL && b == NULL) {
return 0;
}
if (a == NULL) {
return 1;
}
if (b == NULL) {
return -1;
}
if (resource1->sort_index > resource2->sort_index) {
return -1;
}
if (resource1->sort_index < resource2->sort_index) {
return 1;
}
return 0;
}
gint
sort_rsc_priority(gconstpointer a, gconstpointer b)
{
const pe_resource_t *resource1 = (const pe_resource_t *)a;
const pe_resource_t *resource2 = (const pe_resource_t *)b;
if (a == NULL && b == NULL) {
return 0;
}
if (a == NULL) {
return 1;
}
if (b == NULL) {
return -1;
}
if (resource1->priority > resource2->priority) {
return -1;
}
if (resource1->priority < resource2->priority) {
return 1;
}
return 0;
}
static enum pe_quorum_policy
effective_quorum_policy(pe_resource_t *rsc, pe_working_set_t *data_set)
{
enum pe_quorum_policy policy = data_set->no_quorum_policy;
if (is_set(data_set->flags, pe_flag_have_quorum)) {
policy = no_quorum_ignore;
} else if (data_set->no_quorum_policy == no_quorum_demote) {
switch (rsc->role) {
case RSC_ROLE_MASTER:
case RSC_ROLE_SLAVE:
if (rsc->next_role > RSC_ROLE_SLAVE) {
rsc->next_role = RSC_ROLE_SLAVE;
}
policy = no_quorum_ignore;
break;
default:
policy = no_quorum_stop;
break;
}
}
return policy;
}
pe_action_t *
custom_action(pe_resource_t * rsc, char *key, const char *task,
pe_node_t * on_node, gboolean optional, gboolean save_action,
pe_working_set_t * data_set)
{
pe_action_t *action = NULL;
GListPtr possible_matches = NULL;
CRM_CHECK(key != NULL, return NULL);
CRM_CHECK(task != NULL, free(key); return NULL);
if (save_action && rsc != NULL) {
possible_matches = find_actions(rsc->actions, key, on_node);
} else if(save_action) {
#if 0
action = g_hash_table_lookup(data_set->singletons, key);
#else
/* More expensive but takes 'node' into account */
possible_matches = find_actions(data_set->actions, key, on_node);
#endif
}
if(data_set->singletons == NULL) {
data_set->singletons = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, NULL);
}
if (possible_matches != NULL) {
if (pcmk__list_of_multiple(possible_matches)) {
pe_warn("Action %s for %s on %s exists %d times",
task, rsc ? rsc->id : "<NULL>",
on_node ? on_node->details->uname : "<NULL>", g_list_length(possible_matches));
}
action = g_list_nth_data(possible_matches, 0);
pe_rsc_trace(rsc, "Found existing action %d (%s) for %s (%s) on %s",
action->id, action->uuid,
(rsc? rsc->id : "no resource"), task,
(on_node? on_node->details->uname : "no node"));
g_list_free(possible_matches);
}
if (action == NULL) {
if (save_action) {
pe_rsc_trace(rsc, "Creating %s action %d: %s for %s (%s) on %s",
(optional? "optional" : "mandatory"),
data_set->action_id, key,
(rsc? rsc->id : "no resource"), task,
(on_node? on_node->details->uname : "no node"));
}
action = calloc(1, sizeof(pe_action_t));
if (save_action) {
action->id = data_set->action_id++;
} else {
action->id = 0;
}
action->rsc = rsc;
CRM_ASSERT(task != NULL);
action->task = strdup(task);
if (on_node) {
action->node = pe__copy_node(on_node);
}
action->uuid = strdup(key);
if (safe_str_eq(task, CRM_OP_LRM_DELETE)) {
// Resource history deletion for a node can be done on the DC
pe_set_action_bit(action, pe_action_dc);
}
pe_set_action_bit(action, pe_action_runnable);
if (optional) {
pe_set_action_bit(action, pe_action_optional);
} else {
pe_clear_action_bit(action, pe_action_optional);
}
action->extra = crm_str_table_new();
action->meta = crm_str_table_new();
if (save_action) {
data_set->actions = g_list_prepend(data_set->actions, action);
if(rsc == NULL) {
g_hash_table_insert(data_set->singletons, action->uuid, action);
}
}
if (rsc != NULL) {
guint interval_ms = 0;
action->op_entry = find_rsc_op_entry_helper(rsc, key, TRUE);
parse_op_key(key, NULL, NULL, &interval_ms);
unpack_operation(action, action->op_entry, rsc->container, data_set,
interval_ms);
if (save_action) {
rsc->actions = g_list_prepend(rsc->actions, action);
}
}
if (save_action) {
pe_rsc_trace(rsc, "Action %d created", action->id);
}
}
if (!optional && is_set(action->flags, pe_action_optional)) {
pe_rsc_trace(rsc, "Unset optional on action %d", action->id);
pe_clear_action_bit(action, pe_action_optional);
}
if (rsc != NULL) {
enum action_tasks a_task = text2task(action->task);
enum pe_quorum_policy quorum_policy = effective_quorum_policy(rsc, data_set);
int warn_level = LOG_TRACE;
if (save_action) {
warn_level = LOG_WARNING;
}
if (is_set(action->flags, pe_action_have_node_attrs) == FALSE
&& action->node != NULL && action->op_entry != NULL) {
pe_rule_eval_data_t rule_data = {
.node_hash = action->node->details->attrs,
.role = RSC_ROLE_UNKNOWN,
.now = data_set->now,
.match_data = NULL,
.rsc_data = NULL,
.op_data = NULL
};
pe_set_action_bit(action, pe_action_have_node_attrs);
pe__unpack_dataset_nvpairs(action->op_entry, XML_TAG_ATTR_SETS,
&rule_data, action->extra, NULL,
FALSE, data_set);
}
if (is_set(action->flags, pe_action_pseudo)) {
/* leave untouched */
} else if (action->node == NULL) {
pe_rsc_trace(rsc, "Unset runnable on %s", action->uuid);
pe_clear_action_bit(action, pe_action_runnable);
} else if (is_not_set(rsc->flags, pe_rsc_managed)
&& g_hash_table_lookup(action->meta,
XML_LRM_ATTR_INTERVAL_MS) == NULL) {
crm_debug("Action %s (unmanaged)", action->uuid);
pe_rsc_trace(rsc, "Set optional on %s", action->uuid);
pe_set_action_bit(action, pe_action_optional);
/* action->runnable = FALSE; */
} else if (is_not_set(action->flags, pe_action_dc)
&& !(action->node->details->online)
&& (!pe__is_guest_node(action->node)
|| action->node->details->remote_requires_reset)) {
pe_clear_action_bit(action, pe_action_runnable);
do_crm_log(warn_level, "Action %s on %s is unrunnable (offline)",
action->uuid, action->node->details->uname);
if (is_set(action->rsc->flags, pe_rsc_managed)
&& save_action && a_task == stop_rsc
&& action->node->details->unclean == FALSE) {
pe_fence_node(data_set, action->node, "resource actions are unrunnable", FALSE);
}
} else if (is_not_set(action->flags, pe_action_dc)
&& action->node->details->pending) {
pe_clear_action_bit(action, pe_action_runnable);
do_crm_log(warn_level, "Action %s on %s is unrunnable (pending)",
action->uuid, action->node->details->uname);
} else if (action->needs == rsc_req_nothing) {
pe_rsc_trace(rsc, "Action %s does not require anything", action->uuid);
pe_action_set_reason(action, NULL, TRUE);
if (pe__is_guest_node(action->node)
&& !pe_can_fence(data_set, action->node)) {
/* An action that requires nothing usually does not require any
* fencing in order to be runnable. However, there is an
* exception: an action cannot be completed if it is on a guest
* node whose host is unclean and cannot be fenced.
*/
pe_clear_action_bit(action, pe_action_runnable);
crm_debug("%s\t%s (cancelled : host cannot be fenced)",
action->node->details->uname, action->uuid);
} else {
pe_set_action_bit(action, pe_action_runnable);
}
#if 0
/*
* No point checking this
* - if we don't have quorum we can't stonith anyway
*/
} else if (action->needs == rsc_req_stonith) {
crm_trace("Action %s requires only stonith", action->uuid);
action->runnable = TRUE;
#endif
} else if (quorum_policy == no_quorum_stop) {
pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, NULL, "no quorum", pe_action_runnable, TRUE);
crm_debug("%s\t%s (cancelled : quorum)", action->node->details->uname, action->uuid);
} else if (quorum_policy == no_quorum_freeze) {
pe_rsc_trace(rsc, "Check resource is already active: %s %s %s %s", rsc->id, action->uuid, role2text(rsc->next_role), role2text(rsc->role));
if (rsc->fns->active(rsc, TRUE) == FALSE || rsc->next_role > rsc->role) {
pe_action_set_flag_reason(__FUNCTION__, __LINE__, action, NULL, "quorum freeze", pe_action_runnable, TRUE);
pe_rsc_debug(rsc, "%s\t%s (cancelled : quorum freeze)",
action->node->details->uname, action->uuid);
}
} else if(is_not_set(action->flags, pe_action_runnable)) {
pe_rsc_trace(rsc, "Action %s is runnable", action->uuid);
//pe_action_set_reason(action, NULL, TRUE);
pe_set_action_bit(action, pe_action_runnable);
}
if (save_action) {
switch (a_task) {
case stop_rsc:
set_bit(rsc->flags, pe_rsc_stopping);
break;
case start_rsc:
clear_bit(rsc->flags, pe_rsc_starting);
if (is_set(action->flags, pe_action_runnable)) {
set_bit(rsc->flags, pe_rsc_starting);
}
break;
default:
break;
}
}
}
free(key);
return action;
}
static bool
valid_stop_on_fail(const char *value)
{
return pcmk__str_none_of(value, "standby", "demote", "stop", NULL);
}
static const char *
unpack_operation_on_fail(pe_action_t * action)
{
const char *name = NULL;
const char *role = NULL;
const char *on_fail = NULL;
const char *interval_spec = NULL;
const char *enabled = NULL;
const char *value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ON_FAIL);
if (safe_str_eq(action->task, CRMD_ACTION_STOP)
&& !valid_stop_on_fail(value)) {
pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for %s stop "
"action to default value because '%s' is not "
"allowed for stop", action->rsc->id, value);
return NULL;
} else if (safe_str_eq(action->task, CRMD_ACTION_DEMOTE) && !value) {
/* demote on_fail defaults to master monitor value if present */
xmlNode *operation = NULL;
CRM_CHECK(action->rsc != NULL, return NULL);
for (operation = __xml_first_child_element(action->rsc->ops_xml);
operation && !value; operation = __xml_next_element(operation)) {
if (!crm_str_eq((const char *)operation->name, "op", TRUE)) {
continue;
}
name = crm_element_value(operation, "name");
role = crm_element_value(operation, "role");
on_fail = crm_element_value(operation, XML_OP_ATTR_ON_FAIL);
enabled = crm_element_value(operation, "enabled");
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
if (!on_fail) {
continue;
} else if (enabled && !crm_is_true(enabled)) {
continue;
} else if (safe_str_neq(name, "monitor") || safe_str_neq(role, "Master")) {
continue;
} else if (crm_parse_interval_spec(interval_spec) == 0) {
continue;
} else if (safe_str_eq(on_fail, "demote")) {
continue;
}
value = on_fail;
}
} else if (safe_str_eq(action->task, CRM_OP_LRM_DELETE)) {
value = "ignore";
} else if (safe_str_eq(value, "demote")) {
name = crm_element_value(action->op_entry, "name");
role = crm_element_value(action->op_entry, "role");
on_fail = crm_element_value(action->op_entry, XML_OP_ATTR_ON_FAIL);
interval_spec = crm_element_value(action->op_entry,
XML_LRM_ATTR_INTERVAL);
if (safe_str_neq(name, CRMD_ACTION_PROMOTE)
&& (safe_str_neq(name, CRMD_ACTION_STATUS)
|| safe_str_neq(role, "Master")
|| (crm_parse_interval_spec(interval_spec) == 0))) {
pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for %s %s "
"action to default value because 'demote' is not "
"allowed for it", action->rsc->id, name);
return NULL;
}
}
return value;
}
static xmlNode *
find_min_interval_mon(pe_resource_t * rsc, gboolean include_disabled)
{
guint interval_ms = 0;
guint min_interval_ms = G_MAXUINT;
const char *name = NULL;
const char *value = NULL;
const char *interval_spec = NULL;
xmlNode *op = NULL;
xmlNode *operation = NULL;
for (operation = __xml_first_child_element(rsc->ops_xml); operation != NULL;
operation = __xml_next_element(operation)) {
if (crm_str_eq((const char *)operation->name, "op", TRUE)) {
name = crm_element_value(operation, "name");
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
value = crm_element_value(operation, "enabled");
if (!include_disabled && value && crm_is_true(value) == FALSE) {
continue;
}
if (safe_str_neq(name, RSC_STATUS)) {
continue;
}
interval_ms = crm_parse_interval_spec(interval_spec);
if (interval_ms && (interval_ms < min_interval_ms)) {
min_interval_ms = interval_ms;
op = operation;
}
}
}
return op;
}
static int
unpack_start_delay(const char *value, GHashTable *meta)
{
int start_delay = 0;
if (value != NULL) {
start_delay = crm_get_msec(value);
if (start_delay < 0) {
start_delay = 0;
}
if (meta) {
g_hash_table_replace(meta, strdup(XML_OP_ATTR_START_DELAY), crm_itoa(start_delay));
}
}
return start_delay;
}
// true if value contains valid, non-NULL interval origin for recurring op
static bool
unpack_interval_origin(const char *value, xmlNode *xml_obj, guint interval_ms,
crm_time_t *now, long long *start_delay)
{
long long result = 0;
guint interval_sec = interval_ms / 1000;
crm_time_t *origin = NULL;
// Ignore unspecified values and non-recurring operations
if ((value == NULL) || (interval_ms == 0) || (now == NULL)) {
return false;
}
// Parse interval origin from text
origin = crm_time_new(value);
if (origin == NULL) {
pcmk__config_err("Ignoring '" XML_OP_ATTR_ORIGIN "' for operation "
"'%s' because '%s' is not valid",
(ID(xml_obj)? ID(xml_obj) : "(missing ID)"), value);
return false;
}
// Get seconds since origin (negative if origin is in the future)
result = crm_time_get_seconds(now) - crm_time_get_seconds(origin);
crm_time_free(origin);
// Calculate seconds from closest interval to now
result = result % interval_sec;
// Calculate seconds remaining until next interval
result = ((result <= 0)? 0 : interval_sec) - result;
crm_info("Calculated a start delay of %llds for operation '%s'",
result,
(ID(xml_obj)? ID(xml_obj) : "(unspecified)"));
if (start_delay != NULL) {
*start_delay = result * 1000; // milliseconds
}
return true;
}
static int
unpack_timeout(const char *value)
{
int timeout = crm_get_msec(value);
if (timeout < 0) {
timeout = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S);
}
return timeout;
}
int
pe_get_configured_timeout(pe_resource_t *rsc, const char *action, pe_working_set_t *data_set)
{
xmlNode *child = NULL;
const char *timeout = NULL;
int timeout_ms = 0;
pe_rule_eval_data_t rule_data = {
.node_hash = NULL,
.role = RSC_ROLE_UNKNOWN,
.now = data_set->now,
.match_data = NULL,
.rsc_data = NULL,
.op_data = NULL
};
for (child = first_named_child(rsc->ops_xml, XML_ATTR_OP);
child != NULL; child = crm_next_same_xml(child)) {
if (safe_str_eq(action, crm_element_value(child, XML_NVPAIR_ATTR_NAME))) {
timeout = crm_element_value(child, XML_ATTR_TIMEOUT);
break;
}
}
if (timeout == NULL && data_set->op_defaults) {
GHashTable *action_meta = crm_str_table_new();
pe__unpack_dataset_nvpairs(data_set->op_defaults, XML_TAG_META_SETS,
&rule_data, action_meta, NULL, FALSE, data_set);
timeout = g_hash_table_lookup(action_meta, XML_ATTR_TIMEOUT);
}
// @TODO check meta-attributes (including versioned meta-attributes)
// @TODO maybe use min-interval monitor timeout as default for monitors
timeout_ms = crm_get_msec(timeout);
if (timeout_ms < 0) {
timeout_ms = crm_get_msec(CRM_DEFAULT_OP_TIMEOUT_S);
}
return timeout_ms;
}
#if ENABLE_VERSIONED_ATTRS
static void
unpack_versioned_meta(xmlNode *versioned_meta, xmlNode *xml_obj,
guint interval_ms, crm_time_t *now)
{
xmlNode *attrs = NULL;
xmlNode *attr = NULL;
for (attrs = __xml_first_child_element(versioned_meta); attrs != NULL;
attrs = __xml_next_element(attrs)) {
for (attr = __xml_first_child_element(attrs); attr != NULL;
attr = __xml_next_element(attr)) {
const char *name = crm_element_value(attr, XML_NVPAIR_ATTR_NAME);
const char *value = crm_element_value(attr, XML_NVPAIR_ATTR_VALUE);
if (safe_str_eq(name, XML_OP_ATTR_START_DELAY)) {
int start_delay = unpack_start_delay(value, NULL);
crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, start_delay);
} else if (safe_str_eq(name, XML_OP_ATTR_ORIGIN)) {
long long start_delay = 0;
if (unpack_interval_origin(value, xml_obj, interval_ms, now,
&start_delay)) {
crm_xml_add(attr, XML_NVPAIR_ATTR_NAME,
XML_OP_ATTR_START_DELAY);
crm_xml_add_ll(attr, XML_NVPAIR_ATTR_VALUE, start_delay);
}
} else if (safe_str_eq(name, XML_ATTR_TIMEOUT)) {
int timeout = unpack_timeout(value);
crm_xml_add_int(attr, XML_NVPAIR_ATTR_VALUE, timeout);
}
}
}
}
#endif
/*!
* \brief Unpack operation XML into an action structure
*
* Unpack an operation's meta-attributes (normalizing the interval, timeout,
* and start delay values as integer milliseconds), requirements, and
* failure policy.
*
* \param[in,out] action Action to unpack into
* \param[in] xml_obj Operation XML (or NULL if all defaults)
* \param[in] container Resource that contains affected resource, if any
* \param[in] data_set Cluster state
* \param[in] interval_ms How frequently to perform the operation
*/
static void
unpack_operation(pe_action_t * action, xmlNode * xml_obj, pe_resource_t * container,
pe_working_set_t * data_set, guint interval_ms)
{
int timeout = 0;
char *value_ms = NULL;
const char *value = NULL;
const char *field = XML_LRM_ATTR_INTERVAL;
char *default_timeout = NULL;
#if ENABLE_VERSIONED_ATTRS
pe_rsc_action_details_t *rsc_details = NULL;
#endif
pe_rsc_eval_data_t rsc_rule_data = {
.standard = crm_element_value(action->rsc->xml, XML_AGENT_ATTR_CLASS),
.provider = crm_element_value(action->rsc->xml, XML_AGENT_ATTR_PROVIDER),
.agent = crm_element_value(action->rsc->xml, XML_EXPR_ATTR_TYPE)
};
pe_op_eval_data_t op_rule_data = {
.op_name = action->task,
.interval = interval_ms
};
pe_rule_eval_data_t rule_data = {
.node_hash = NULL,
.role = RSC_ROLE_UNKNOWN,
.now = data_set->now,
.match_data = NULL,
.rsc_data = &rsc_rule_data,
.op_data = &op_rule_data
};
CRM_CHECK(action && action->rsc, return);
// Cluster-wide <op_defaults> <meta_attributes>
pe__unpack_dataset_nvpairs(data_set->op_defaults, XML_TAG_META_SETS, &rule_data,
action->meta, NULL, FALSE, data_set);
// Probe timeouts default differently, so handle timeout default later
default_timeout = g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT);
if (default_timeout) {
default_timeout = strdup(default_timeout);
g_hash_table_remove(action->meta, XML_ATTR_TIMEOUT);
}
if (xml_obj) {
xmlAttrPtr xIter = NULL;
// <op> <meta_attributes> take precedence over defaults
pe__unpack_dataset_nvpairs(xml_obj, XML_TAG_META_SETS, &rule_data,
action->meta, NULL, TRUE, data_set);
#if ENABLE_VERSIONED_ATTRS
rsc_details = pe_rsc_action_details(action);
pe_eval_versioned_attributes(data_set->input, xml_obj,
XML_TAG_ATTR_SETS, &rule_data,
rsc_details->versioned_parameters,
NULL);
pe_eval_versioned_attributes(data_set->input, xml_obj,
XML_TAG_META_SETS, &rule_data,
rsc_details->versioned_meta,
NULL);
#endif
/* Anything set as an <op> XML property has highest precedence.
* This ensures we use the name and interval from the <op> tag.
*/
for (xIter = xml_obj->properties; xIter; xIter = xIter->next) {
const char *prop_name = (const char *)xIter->name;
const char *prop_value = crm_element_value(xml_obj, prop_name);
g_hash_table_replace(action->meta, strdup(prop_name), strdup(prop_value));
}
}
g_hash_table_remove(action->meta, "id");
// Normalize interval to milliseconds
if (interval_ms > 0) {
value_ms = crm_strdup_printf("%u", interval_ms);
g_hash_table_replace(action->meta, strdup(field), value_ms);
} else if (g_hash_table_lookup(action->meta, field) != NULL) {
g_hash_table_remove(action->meta, field);
}
// Handle timeout default, now that we know the interval
if (g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT)) {
free(default_timeout);
} else {
// Probe timeouts default to minimum-interval monitor's
if (safe_str_eq(action->task, RSC_STATUS) && (interval_ms == 0)) {
xmlNode *min_interval_mon = find_min_interval_mon(action->rsc, FALSE);
if (min_interval_mon) {
value = crm_element_value(min_interval_mon, XML_ATTR_TIMEOUT);
if (value) {
crm_trace("\t%s defaults to minimum-interval monitor's timeout '%s'",
action->uuid, value);
free(default_timeout);
default_timeout = strdup(value);
}
}
}
if (default_timeout) {
g_hash_table_insert(action->meta, strdup(XML_ATTR_TIMEOUT),
default_timeout);
}
}
if (pcmk__str_none_of(action->task, RSC_START, RSC_PROMOTE, NULL)) {
action->needs = rsc_req_nothing;
value = "nothing (not start/promote)";
} else if (is_set(action->rsc->flags, pe_rsc_needs_fencing)) {
action->needs = rsc_req_stonith;
value = "fencing (resource)";
} else if (is_set(action->rsc->flags, pe_rsc_needs_quorum)) {
action->needs = rsc_req_quorum;
value = "quorum (resource)";
} else {
action->needs = rsc_req_nothing;
value = "nothing (resource)";
}
pe_rsc_trace(action->rsc, "\tAction %s requires: %s", action->uuid, value);
value = unpack_operation_on_fail(action);
if (value == NULL) {
} else if (safe_str_eq(value, "block")) {
action->on_fail = action_fail_block;
g_hash_table_insert(action->meta, strdup(XML_OP_ATTR_ON_FAIL), strdup("block"));
value = "block"; // The above could destroy the original string
} else if (safe_str_eq(value, "fence")) {
action->on_fail = action_fail_fence;
value = "node fencing";
if (is_set(data_set->flags, pe_flag_stonith_enabled) == FALSE) {
pcmk__config_err("Resetting '" XML_OP_ATTR_ON_FAIL "' for "
"operation '%s' to 'stop' because 'fence' is not "
"valid when fencing is disabled", action->uuid);
action->on_fail = action_fail_stop;
action->fail_role = RSC_ROLE_STOPPED;
value = "stop resource";
}
} else if (safe_str_eq(value, "standby")) {
action->on_fail = action_fail_standby;
value = "node standby";
} else if (pcmk__str_any_of(value, "ignore", "nothing", NULL)) {
action->on_fail = action_fail_ignore;
value = "ignore";
} else if (safe_str_eq(value, "migrate")) {
action->on_fail = action_fail_migrate;
value = "force migration";
} else if (safe_str_eq(value, "stop")) {
action->on_fail = action_fail_stop;
action->fail_role = RSC_ROLE_STOPPED;
value = "stop resource";
} else if (safe_str_eq(value, "restart")) {
action->on_fail = action_fail_recover;
value = "restart (and possibly migrate)";
} else if (safe_str_eq(value, "restart-container")) {
if (container) {
action->on_fail = action_fail_restart_container;
value = "restart container (and possibly migrate)";
} else {
value = NULL;
}
} else if (safe_str_eq(value, "demote")) {
action->on_fail = action_fail_demote;
value = "demote instance";
} else {
pe_err("Resource %s: Unknown failure type (%s)", action->rsc->id, value);
value = NULL;
}
/* defaults */
if (value == NULL && container) {
action->on_fail = action_fail_restart_container;
value = "restart container (and possibly migrate) (default)";
/* For remote nodes, ensure that any failure that results in dropping an
* active connection to the node results in fencing of the node.
*
* There are only two action failures that don't result in fencing.
* 1. probes - probe failures are expected.
* 2. start - a start failure indicates that an active connection does not already
* exist. The user can set op on-fail=fence if they really want to fence start
* failures. */
} else if (((value == NULL) || !is_set(action->rsc->flags, pe_rsc_managed)) &&
(pe__resource_is_remote_conn(action->rsc, data_set) &&
!(safe_str_eq(action->task, CRMD_ACTION_STATUS) && (interval_ms == 0)) &&
(safe_str_neq(action->task, CRMD_ACTION_START)))) {
if (!is_set(action->rsc->flags, pe_rsc_managed)) {
action->on_fail = action_fail_stop;
action->fail_role = RSC_ROLE_STOPPED;
value = "stop unmanaged remote node (enforcing default)";
} else {
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
value = "fence remote node (default)";
} else {
value = "recover remote node connection (default)";
}
if (action->rsc->remote_reconnect_ms) {
action->fail_role = RSC_ROLE_STOPPED;
}
action->on_fail = action_fail_reset_remote;
}
} else if (value == NULL && safe_str_eq(action->task, CRMD_ACTION_STOP)) {
if (is_set(data_set->flags, pe_flag_stonith_enabled)) {
action->on_fail = action_fail_fence;
value = "resource fence (default)";
} else {
action->on_fail = action_fail_block;
value = "resource block (default)";
}
} else if (value == NULL) {
action->on_fail = action_fail_recover;
value = "restart (and possibly migrate) (default)";
}
pe_rsc_trace(action->rsc, "\t%s failure handling: %s", action->task, value);
value = NULL;
if (xml_obj != NULL) {
value = g_hash_table_lookup(action->meta, "role_after_failure");
if (value) {
pe_warn_once(pe_wo_role_after,
"Support for role_after_failure is deprecated and will be removed in a future release");
}
}
if (value != NULL && action->fail_role == RSC_ROLE_UNKNOWN) {
action->fail_role = text2role(value);
}
/* defaults */
if (action->fail_role == RSC_ROLE_UNKNOWN) {
if (safe_str_eq(action->task, CRMD_ACTION_PROMOTE)) {
action->fail_role = RSC_ROLE_SLAVE;
} else {
action->fail_role = RSC_ROLE_STARTED;
}
}
pe_rsc_trace(action->rsc, "\t%s failure results in: %s", action->task,
role2text(action->fail_role));
value = g_hash_table_lookup(action->meta, XML_OP_ATTR_START_DELAY);
if (value) {
unpack_start_delay(value, action->meta);
} else {
long long start_delay = 0;
value = g_hash_table_lookup(action->meta, XML_OP_ATTR_ORIGIN);
if (unpack_interval_origin(value, xml_obj, interval_ms, data_set->now,
&start_delay)) {
g_hash_table_replace(action->meta, strdup(XML_OP_ATTR_START_DELAY),
crm_strdup_printf("%lld", start_delay));
}
}
value = g_hash_table_lookup(action->meta, XML_ATTR_TIMEOUT);
timeout = unpack_timeout(value);
g_hash_table_replace(action->meta, strdup(XML_ATTR_TIMEOUT), crm_itoa(timeout));
#if ENABLE_VERSIONED_ATTRS
unpack_versioned_meta(rsc_details->versioned_meta, xml_obj, interval_ms,
data_set->now);
#endif
}
static xmlNode *
find_rsc_op_entry_helper(pe_resource_t * rsc, const char *key, gboolean include_disabled)
{
guint interval_ms = 0;
gboolean do_retry = TRUE;
char *local_key = NULL;
const char *name = NULL;
const char *value = NULL;
const char *interval_spec = NULL;
char *match_key = NULL;
xmlNode *op = NULL;
xmlNode *operation = NULL;
retry:
for (operation = __xml_first_child_element(rsc->ops_xml); operation != NULL;
operation = __xml_next_element(operation)) {
if (crm_str_eq((const char *)operation->name, "op", TRUE)) {
name = crm_element_value(operation, "name");
interval_spec = crm_element_value(operation, XML_LRM_ATTR_INTERVAL);
value = crm_element_value(operation, "enabled");
if (!include_disabled && value && crm_is_true(value) == FALSE) {
continue;
}
interval_ms = crm_parse_interval_spec(interval_spec);
match_key = pcmk__op_key(rsc->id, name, interval_ms);
if (safe_str_eq(key, match_key)) {
op = operation;
}
free(match_key);
if (rsc->clone_name) {
match_key = pcmk__op_key(rsc->clone_name, name, interval_ms);
if (safe_str_eq(key, match_key)) {
op = operation;
}
free(match_key);
}
if (op != NULL) {
free(local_key);
return op;
}
}
}
free(local_key);
if (do_retry == FALSE) {
return NULL;
}
do_retry = FALSE;
if (strstr(key, CRMD_ACTION_MIGRATE) || strstr(key, CRMD_ACTION_MIGRATED)) {
local_key = pcmk__op_key(rsc->id, "migrate", 0);
key = local_key;
goto retry;
} else if (strstr(key, "_notify_")) {
local_key = pcmk__op_key(rsc->id, "notify", 0);
key = local_key;
goto retry;
}
return NULL;
}
xmlNode *
find_rsc_op_entry(pe_resource_t * rsc, const char *key)
{
return find_rsc_op_entry_helper(rsc, key, FALSE);
}
void
print_node(const char *pre_text, pe_node_t * node, gboolean details)
{
if (node == NULL) {
crm_trace("%s%s: <NULL>", pre_text == NULL ? "" : pre_text, pre_text == NULL ? "" : ": ");
return;
}
CRM_ASSERT(node->details);
crm_trace("%s%s%sNode %s: (weight=%d, fixed=%s)",
pre_text == NULL ? "" : pre_text,
pre_text == NULL ? "" : ": ",
node->details->online ? "" : "Unavailable/Unclean ",
node->details->uname, node->weight, node->fixed ? "True" : "False");
if (details) {
int log_level = LOG_TRACE;
char *pe_mutable = strdup("\t\t");
GListPtr gIter = node->details->running_rsc;
crm_trace("\t\t===Node Attributes");
g_hash_table_foreach(node->details->attrs, print_str_str, pe_mutable);
free(pe_mutable);
crm_trace("\t\t=== Resources");
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *rsc = (pe_resource_t *) gIter->data;
rsc->fns->print(rsc, "\t\t", pe_print_log|pe_print_pending,
&log_level);
}
}
}
/*
* Used by the HashTable for-loop
*/
void
print_str_str(gpointer key, gpointer value, gpointer user_data)
{
crm_trace("%s%s %s ==> %s",
user_data == NULL ? "" : (char *)user_data,
user_data == NULL ? "" : ": ", (char *)key, (char *)value);
}
void
pe_free_action(pe_action_t * action)
{
if (action == NULL) {
return;
}
g_list_free_full(action->actions_before, free); /* pe_action_wrapper_t* */
g_list_free_full(action->actions_after, free); /* pe_action_wrapper_t* */
if (action->extra) {
g_hash_table_destroy(action->extra);
}
if (action->meta) {
g_hash_table_destroy(action->meta);
}
#if ENABLE_VERSIONED_ATTRS
if (action->rsc) {
pe_free_rsc_action_details(action);
}
#endif
free(action->cancel_task);
free(action->reason);
free(action->task);
free(action->uuid);
free(action->node);
free(action);
}
GListPtr
find_recurring_actions(GListPtr input, pe_node_t * not_on_node)
{
const char *value = NULL;
GListPtr result = NULL;
GListPtr gIter = input;
CRM_CHECK(input != NULL, return NULL);
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
value = g_hash_table_lookup(action->meta, XML_LRM_ATTR_INTERVAL_MS);
if (value == NULL) {
/* skip */
} else if (safe_str_eq(value, "0")) {
/* skip */
} else if (safe_str_eq(CRMD_ACTION_CANCEL, action->task)) {
/* skip */
} else if (not_on_node == NULL) {
crm_trace("(null) Found: %s", action->uuid);
result = g_list_prepend(result, action);
} else if (action->node == NULL) {
/* skip */
} else if (action->node->details != not_on_node->details) {
crm_trace("Found: %s", action->uuid);
result = g_list_prepend(result, action);
}
}
return result;
}
enum action_tasks
get_complex_task(pe_resource_t * rsc, const char *name, gboolean allow_non_atomic)
{
enum action_tasks task = text2task(name);
if (rsc == NULL) {
return task;
} else if (allow_non_atomic == FALSE || rsc->variant == pe_native) {
switch (task) {
case stopped_rsc:
case started_rsc:
case action_demoted:
case action_promoted:
crm_trace("Folding %s back into its atomic counterpart for %s", name, rsc->id);
return task - 1;
- break;
default:
break;
}
}
return task;
}
pe_action_t *
find_first_action(GListPtr input, const char *uuid, const char *task, pe_node_t * on_node)
{
GListPtr gIter = NULL;
CRM_CHECK(uuid || task, return NULL);
for (gIter = input; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (uuid != NULL && safe_str_neq(uuid, action->uuid)) {
continue;
} else if (task != NULL && safe_str_neq(task, action->task)) {
continue;
} else if (on_node == NULL) {
return action;
} else if (action->node == NULL) {
continue;
} else if (on_node->details == action->node->details) {
return action;
}
}
return NULL;
}
GListPtr
find_actions(GListPtr input, const char *key, const pe_node_t *on_node)
{
GListPtr gIter = input;
GListPtr result = NULL;
CRM_CHECK(key != NULL, return NULL);
for (; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (safe_str_neq(key, action->uuid)) {
crm_trace("%s does not match action %s", key, action->uuid);
continue;
} else if (on_node == NULL) {
crm_trace("Action %s matches (ignoring node)", key);
result = g_list_prepend(result, action);
} else if (action->node == NULL) {
crm_trace("Action %s matches (unallocated, assigning to %s)",
key, on_node->details->uname);
action->node = pe__copy_node(on_node);
result = g_list_prepend(result, action);
} else if (on_node->details == action->node->details) {
crm_trace("Action %s on %s matches", key, on_node->details->uname);
result = g_list_prepend(result, action);
} else {
crm_trace("Action %s on node %s does not match requested node %s",
key, action->node->details->uname,
on_node->details->uname);
}
}
return result;
}
GList *
find_actions_exact(GList *input, const char *key, const pe_node_t *on_node)
{
GList *result = NULL;
CRM_CHECK(key != NULL, return NULL);
if (on_node == NULL) {
crm_trace("Not searching for action %s because node not specified",
key);
return NULL;
}
for (GList *gIter = input; gIter != NULL; gIter = gIter->next) {
pe_action_t *action = (pe_action_t *) gIter->data;
if (action->node == NULL) {
crm_trace("Skipping comparison of %s vs action %s without node",
key, action->uuid);
} else if (safe_str_neq(key, action->uuid)) {
crm_trace("Desired action %s doesn't match %s", key, action->uuid);
} else if (safe_str_neq(on_node->details->id,
action->node->details->id)) {
crm_trace("Action %s desired node ID %s doesn't match %s",
key, on_node->details->id, action->node->details->id);
} else {
crm_trace("Action %s matches", key);
result = g_list_prepend(result, action);
}
}
return result;
}
/*!
* \brief Find all actions of given type for a resource
*
* \param[in] rsc Resource to search
* \param[in] node Find only actions scheduled on this node
* \param[in] task Action name to search for
* \param[in] require_node If TRUE, NULL node or action node will not match
*
* \return List of actions found (or NULL if none)
* \note If node is not NULL and require_node is FALSE, matching actions
* without a node will be assigned to node.
*/
GList *
pe__resource_actions(const pe_resource_t *rsc, const pe_node_t *node,
const char *task, bool require_node)
{
GList *result = NULL;
char *key = pcmk__op_key(rsc->id, task, 0);
if (require_node) {
result = find_actions_exact(rsc->actions, key, node);
} else {
result = find_actions(rsc->actions, key, node);
}
free(key);
return result;
}
static void
resource_node_score(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag)
{
pe_node_t *match = NULL;
if ((rsc->exclusive_discover || (node->rsc_discover_mode == pe_discover_never))
&& safe_str_eq(tag, "symmetric_default")) {
/* This string comparision may be fragile, but exclusive resources and
* exclusive nodes should not have the symmetric_default constraint
* applied to them.
*/
return;
} else if (rsc->children) {
GListPtr gIter = rsc->children;
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
resource_node_score(child_rsc, node, score, tag);
}
}
pe_rsc_trace(rsc, "Setting %s for %s on %s: %d", tag, rsc->id, node->details->uname, score);
match = pe_hash_table_lookup(rsc->allowed_nodes, node->details->id);
if (match == NULL) {
match = pe__copy_node(node);
g_hash_table_insert(rsc->allowed_nodes, (gpointer) match->details->id, match);
}
match->weight = pe__add_scores(match->weight, score);
}
void
resource_location(pe_resource_t * rsc, pe_node_t * node, int score, const char *tag,
pe_working_set_t * data_set)
{
if (node != NULL) {
resource_node_score(rsc, node, score, tag);
} else if (data_set != NULL) {
GListPtr gIter = data_set->nodes;
for (; gIter != NULL; gIter = gIter->next) {
pe_node_t *node_iter = (pe_node_t *) gIter->data;
resource_node_score(rsc, node_iter, score, tag);
}
} else {
GHashTableIter iter;
pe_node_t *node_iter = NULL;
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node_iter)) {
resource_node_score(rsc, node_iter, score, tag);
}
}
if (node == NULL && score == -INFINITY) {
if (rsc->allocated_to) {
crm_info("Deallocating %s from %s", rsc->id, rsc->allocated_to->details->uname);
free(rsc->allocated_to);
rsc->allocated_to = NULL;
}
}
}
#define sort_return(an_int, why) do { \
free(a_uuid); \
free(b_uuid); \
crm_trace("%s (%d) %c %s (%d) : %s", \
a_xml_id, a_call_id, an_int>0?'>':an_int<0?'<':'=', \
b_xml_id, b_call_id, why); \
return an_int; \
} while(0)
gint
sort_op_by_callid(gconstpointer a, gconstpointer b)
{
int a_call_id = -1;
int b_call_id = -1;
char *a_uuid = NULL;
char *b_uuid = NULL;
const xmlNode *xml_a = a;
const xmlNode *xml_b = b;
const char *a_xml_id = crm_element_value(xml_a, XML_ATTR_ID);
const char *b_xml_id = crm_element_value(xml_b, XML_ATTR_ID);
if (safe_str_eq(a_xml_id, b_xml_id)) {
/* We have duplicate lrm_rsc_op entries in the status
* section which is unlikely to be a good thing
* - we can handle it easily enough, but we need to get
* to the bottom of why it's happening.
*/
pe_err("Duplicate lrm_rsc_op entries named %s", a_xml_id);
sort_return(0, "duplicate");
}
crm_element_value_int(xml_a, XML_LRM_ATTR_CALLID, &a_call_id);
crm_element_value_int(xml_b, XML_LRM_ATTR_CALLID, &b_call_id);
if (a_call_id == -1 && b_call_id == -1) {
/* both are pending ops so it doesn't matter since
* stops are never pending
*/
sort_return(0, "pending");
} else if (a_call_id >= 0 && a_call_id < b_call_id) {
sort_return(-1, "call id");
} else if (b_call_id >= 0 && a_call_id > b_call_id) {
sort_return(1, "call id");
} else if (b_call_id >= 0 && a_call_id == b_call_id) {
/*
* The op and last_failed_op are the same
* Order on last-rc-change
*/
time_t last_a = -1;
time_t last_b = -1;
crm_element_value_epoch(xml_a, XML_RSC_OP_LAST_CHANGE, &last_a);
crm_element_value_epoch(xml_b, XML_RSC_OP_LAST_CHANGE, &last_b);
crm_trace("rc-change: %lld vs %lld",
(long long) last_a, (long long) last_b);
if (last_a >= 0 && last_a < last_b) {
sort_return(-1, "rc-change");
} else if (last_b >= 0 && last_a > last_b) {
sort_return(1, "rc-change");
}
sort_return(0, "rc-change");
} else {
/* One of the inputs is a pending operation
* Attempt to use XML_ATTR_TRANSITION_MAGIC to determine its age relative to the other
*/
int a_id = -1;
int b_id = -1;
const char *a_magic = crm_element_value(xml_a, XML_ATTR_TRANSITION_MAGIC);
const char *b_magic = crm_element_value(xml_b, XML_ATTR_TRANSITION_MAGIC);
CRM_CHECK(a_magic != NULL && b_magic != NULL, sort_return(0, "No magic"));
if (!decode_transition_magic(a_magic, &a_uuid, &a_id, NULL, NULL, NULL,
NULL)) {
sort_return(0, "bad magic a");
}
if (!decode_transition_magic(b_magic, &b_uuid, &b_id, NULL, NULL, NULL,
NULL)) {
sort_return(0, "bad magic b");
}
/* try to determine the relative age of the operation...
* some pending operations (e.g. a start) may have been superseded
* by a subsequent stop
*
* [a|b]_id == -1 means it's a shutdown operation and _always_ comes last
*/
if (safe_str_neq(a_uuid, b_uuid) || a_id == b_id) {
/*
* some of the logic in here may be redundant...
*
* if the UUID from the TE doesn't match then one better
* be a pending operation.
* pending operations don't survive between elections and joins
* because we query the LRM directly
*/
if (b_call_id == -1) {
sort_return(-1, "transition + call");
} else if (a_call_id == -1) {
sort_return(1, "transition + call");
}
} else if ((a_id >= 0 && a_id < b_id) || b_id == -1) {
sort_return(-1, "transition");
} else if ((b_id >= 0 && a_id > b_id) || a_id == -1) {
sort_return(1, "transition");
}
}
/* we should never end up here */
CRM_CHECK(FALSE, sort_return(0, "default"));
}
time_t
get_effective_time(pe_working_set_t * data_set)
{
if(data_set) {
if (data_set->now == NULL) {
crm_trace("Recording a new 'now'");
data_set->now = crm_time_new(NULL);
}
return crm_time_get_seconds_since_epoch(data_set->now);
}
crm_trace("Defaulting to 'now'");
return time(NULL);
}
gboolean
get_target_role(pe_resource_t * rsc, enum rsc_role_e * role)
{
enum rsc_role_e local_role = RSC_ROLE_UNKNOWN;
const char *value = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
CRM_CHECK(role != NULL, return FALSE);
if (value == NULL || safe_str_eq("started", value)
|| safe_str_eq("default", value)) {
return FALSE;
}
local_role = text2role(value);
if (local_role == RSC_ROLE_UNKNOWN) {
pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s "
"because '%s' is not valid", rsc->id, value);
return FALSE;
} else if (local_role > RSC_ROLE_STARTED) {
if (is_set(uber_parent(rsc)->flags, pe_rsc_promotable)) {
if (local_role > RSC_ROLE_SLAVE) {
/* This is what we'd do anyway, just leave the default to avoid messing up the placement algorithm */
return FALSE;
}
} else {
pcmk__config_err("Ignoring '" XML_RSC_ATTR_TARGET_ROLE "' for %s "
"because '%s' only makes sense for promotable "
"clones", rsc->id, value);
return FALSE;
}
}
*role = local_role;
return TRUE;
}
gboolean
order_actions(pe_action_t * lh_action, pe_action_t * rh_action, enum pe_ordering order)
{
GListPtr gIter = NULL;
pe_action_wrapper_t *wrapper = NULL;
GListPtr list = NULL;
if (order == pe_order_none) {
return FALSE;
}
if (lh_action == NULL || rh_action == NULL) {
return FALSE;
}
crm_trace("Ordering Action %s before %s", lh_action->uuid, rh_action->uuid);
/* Ensure we never create a dependency on ourselves... it's happened */
CRM_ASSERT(lh_action != rh_action);
/* Filter dups, otherwise update_action_states() has too much work to do */
gIter = lh_action->actions_after;
for (; gIter != NULL; gIter = gIter->next) {
pe_action_wrapper_t *after = (pe_action_wrapper_t *) gIter->data;
if (after->action == rh_action && (after->type & order)) {
return FALSE;
}
}
wrapper = calloc(1, sizeof(pe_action_wrapper_t));
wrapper->action = rh_action;
wrapper->type = order;
list = lh_action->actions_after;
list = g_list_prepend(list, wrapper);
lh_action->actions_after = list;
wrapper = NULL;
/* order |= pe_order_implies_then; */
/* order ^= pe_order_implies_then; */
wrapper = calloc(1, sizeof(pe_action_wrapper_t));
wrapper->action = lh_action;
wrapper->type = order;
list = rh_action->actions_before;
list = g_list_prepend(list, wrapper);
rh_action->actions_before = list;
return TRUE;
}
pe_action_t *
get_pseudo_op(const char *name, pe_working_set_t * data_set)
{
pe_action_t *op = NULL;
if(data_set->singletons) {
op = g_hash_table_lookup(data_set->singletons, name);
}
if (op == NULL) {
op = custom_action(NULL, strdup(name), name, NULL, TRUE, TRUE, data_set);
set_bit(op->flags, pe_action_pseudo);
set_bit(op->flags, pe_action_runnable);
}
return op;
}
void
destroy_ticket(gpointer data)
{
pe_ticket_t *ticket = data;
if (ticket->state) {
g_hash_table_destroy(ticket->state);
}
free(ticket->id);
free(ticket);
}
pe_ticket_t *
ticket_new(const char *ticket_id, pe_working_set_t * data_set)
{
pe_ticket_t *ticket = NULL;
if (ticket_id == NULL || strlen(ticket_id) == 0) {
return NULL;
}
if (data_set->tickets == NULL) {
data_set->tickets =
g_hash_table_new_full(crm_str_hash, g_str_equal, free,
destroy_ticket);
}
ticket = g_hash_table_lookup(data_set->tickets, ticket_id);
if (ticket == NULL) {
ticket = calloc(1, sizeof(pe_ticket_t));
if (ticket == NULL) {
crm_err("Cannot allocate ticket '%s'", ticket_id);
return NULL;
}
crm_trace("Creaing ticket entry for %s", ticket_id);
ticket->id = strdup(ticket_id);
ticket->granted = FALSE;
ticket->last_granted = -1;
ticket->standby = FALSE;
ticket->state = crm_str_table_new();
g_hash_table_insert(data_set->tickets, strdup(ticket->id), ticket);
}
return ticket;
}
static void
filter_parameters(xmlNode * param_set, const char *param_string, bool need_present)
{
if (param_set && param_string) {
xmlAttrPtr xIter = param_set->properties;
while (xIter) {
const char *prop_name = (const char *)xIter->name;
char *name = crm_strdup_printf(" %s ", prop_name);
char *match = strstr(param_string, name);
free(name);
// Do now, because current entry might get removed below
xIter = xIter->next;
if (need_present && match == NULL) {
crm_trace("%s not found in %s", prop_name, param_string);
xml_remove_prop(param_set, prop_name);
} else if (need_present == FALSE && match) {
crm_trace("%s found in %s", prop_name, param_string);
xml_remove_prop(param_set, prop_name);
}
}
}
}
#if ENABLE_VERSIONED_ATTRS
static void
append_versioned_params(xmlNode *versioned_params, const char *ra_version, xmlNode *params)
{
GHashTable *hash = pe_unpack_versioned_parameters(versioned_params, ra_version);
char *key = NULL;
char *value = NULL;
GHashTableIter iter;
g_hash_table_iter_init(&iter, hash);
while (g_hash_table_iter_next(&iter, (gpointer *) &key, (gpointer *) &value)) {
crm_xml_add(params, key, value);
}
g_hash_table_destroy(hash);
}
#endif
/*!
* \internal
* \brief Calculate action digests and store in node's digest cache
*
* \param[in] rsc Resource that action was for
* \param[in] task Name of action performed
* \param[in] key Action's task key
* \param[in] node Node action was performed on
* \param[in] xml_op XML of operation in CIB status (if available)
* \param[in] calc_secure Whether to calculate secure digest
* \param[in] data_set Cluster working set
*
* \return Pointer to node's digest cache entry
*/
static op_digest_cache_t *
rsc_action_digest(pe_resource_t *rsc, const char *task, const char *key,
pe_node_t *node, xmlNode *xml_op, bool calc_secure,
pe_working_set_t *data_set)
{
op_digest_cache_t *data = NULL;
data = g_hash_table_lookup(node->details->digest_cache, key);
if (data == NULL) {
GHashTable *local_rsc_params = crm_str_table_new();
pe_action_t *action = custom_action(rsc, strdup(key), task, node, TRUE, FALSE, data_set);
#if ENABLE_VERSIONED_ATTRS
xmlNode *local_versioned_params = create_xml_node(NULL, XML_TAG_RSC_VER_ATTRS);
const char *ra_version = NULL;
#endif
const char *op_version;
const char *restart_list = NULL;
const char *secure_list = " passwd password ";
data = calloc(1, sizeof(op_digest_cache_t));
CRM_ASSERT(data != NULL);
get_rsc_attributes(local_rsc_params, rsc, node, data_set);
#if ENABLE_VERSIONED_ATTRS
pe_get_versioned_attributes(local_versioned_params, rsc, node, data_set);
#endif
data->params_all = create_xml_node(NULL, XML_TAG_PARAMS);
// REMOTE_CONTAINER_HACK: Allow remote nodes that start containers with pacemaker remote inside
if (pe__add_bundle_remote_name(rsc, data->params_all,
XML_RSC_ATTR_REMOTE_RA_ADDR)) {
crm_trace("Set address for bundle connection %s (on %s)",
rsc->id, node->details->uname);
}
g_hash_table_foreach(local_rsc_params, hash2field, data->params_all);
g_hash_table_foreach(action->extra, hash2field, data->params_all);
g_hash_table_foreach(rsc->parameters, hash2field, data->params_all);
g_hash_table_foreach(action->meta, hash2metafield, data->params_all);
if(xml_op) {
secure_list = crm_element_value(xml_op, XML_LRM_ATTR_OP_SECURE);
restart_list = crm_element_value(xml_op, XML_LRM_ATTR_OP_RESTART);
op_version = crm_element_value(xml_op, XML_ATTR_CRM_VERSION);
#if ENABLE_VERSIONED_ATTRS
ra_version = crm_element_value(xml_op, XML_ATTR_RA_VERSION);
#endif
} else {
op_version = CRM_FEATURE_SET;
}
#if ENABLE_VERSIONED_ATTRS
append_versioned_params(local_versioned_params, ra_version, data->params_all);
append_versioned_params(rsc->versioned_parameters, ra_version, data->params_all);
{
pe_rsc_action_details_t *details = pe_rsc_action_details(action);
append_versioned_params(details->versioned_parameters, ra_version, data->params_all);
}
#endif
pcmk__filter_op_for_digest(data->params_all);
g_hash_table_destroy(local_rsc_params);
pe_free_action(action);
data->digest_all_calc = calculate_operation_digest(data->params_all, op_version);
if (calc_secure) {
data->params_secure = copy_xml(data->params_all);
if(secure_list) {
filter_parameters(data->params_secure, secure_list, FALSE);
}
data->digest_secure_calc = calculate_operation_digest(data->params_secure, op_version);
}
if(xml_op && crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST) != NULL) {
data->params_restart = copy_xml(data->params_all);
if (restart_list) {
filter_parameters(data->params_restart, restart_list, TRUE);
}
data->digest_restart_calc = calculate_operation_digest(data->params_restart, op_version);
}
g_hash_table_insert(node->details->digest_cache, strdup(key), data);
}
return data;
}
op_digest_cache_t *
rsc_action_digest_cmp(pe_resource_t * rsc, xmlNode * xml_op, pe_node_t * node,
pe_working_set_t * data_set)
{
op_digest_cache_t *data = NULL;
char *key = NULL;
guint interval_ms = 0;
const char *op_version;
const char *task = crm_element_value(xml_op, XML_LRM_ATTR_TASK);
const char *digest_all;
const char *digest_restart;
CRM_ASSERT(node != NULL);
op_version = crm_element_value(xml_op, XML_ATTR_CRM_VERSION);
digest_all = crm_element_value(xml_op, XML_LRM_ATTR_OP_DIGEST);
digest_restart = crm_element_value(xml_op, XML_LRM_ATTR_RESTART_DIGEST);
crm_element_value_ms(xml_op, XML_LRM_ATTR_INTERVAL_MS, &interval_ms);
key = pcmk__op_key(rsc->id, task, interval_ms);
data = rsc_action_digest(rsc, task, key, node, xml_op,
is_set(data_set->flags, pe_flag_sanitized),
data_set);
data->rc = RSC_DIGEST_MATCH;
if (digest_restart && data->digest_restart_calc && strcmp(data->digest_restart_calc, digest_restart) != 0) {
pe_rsc_info(rsc, "Parameters to %s on %s changed: was %s vs. now %s (restart:%s) %s",
key, node->details->uname,
crm_str(digest_restart), data->digest_restart_calc,
op_version, crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC));
data->rc = RSC_DIGEST_RESTART;
} else if (digest_all == NULL) {
/* it is unknown what the previous op digest was */
data->rc = RSC_DIGEST_UNKNOWN;
} else if (strcmp(digest_all, data->digest_all_calc) != 0) {
pe_rsc_info(rsc, "Parameters to %s on %s changed: was %s vs. now %s (%s:%s) %s",
key, node->details->uname,
crm_str(digest_all), data->digest_all_calc,
(interval_ms > 0)? "reschedule" : "reload",
op_version, crm_element_value(xml_op, XML_ATTR_TRANSITION_MAGIC));
data->rc = RSC_DIGEST_ALL;
}
free(key);
return data;
}
/*!
* \internal
* \brief Create an unfencing summary for use in special node attribute
*
* Create a string combining a fence device's resource ID, agent type, and
* parameter digest (whether for all parameters or just non-private parameters).
* This can be stored in a special node attribute, allowing us to detect changes
* in either the agent type or parameters, to know whether unfencing must be
* redone or can be safely skipped when the device's history is cleaned.
*
* \param[in] rsc_id Fence device resource ID
* \param[in] agent_type Fence device agent
* \param[in] param_digest Fence device parameter digest
*
* \return Newly allocated string with unfencing digest
* \note The caller is responsible for freeing the result.
*/
static inline char *
create_unfencing_summary(const char *rsc_id, const char *agent_type,
const char *param_digest)
{
return crm_strdup_printf("%s:%s:%s", rsc_id, agent_type, param_digest);
}
/*!
* \internal
* \brief Check whether a node can skip unfencing
*
* Check whether a fence device's current definition matches a node's
* stored summary of when it was last unfenced by the device.
*
* \param[in] rsc_id Fence device's resource ID
* \param[in] agent Fence device's agent type
* \param[in] digest_calc Fence device's current parameter digest
* \param[in] node_summary Value of node's special unfencing node attribute
* (a comma-separated list of unfencing summaries for
* all devices that have unfenced this node)
*
* \return TRUE if digest matches, FALSE otherwise
*/
static bool
unfencing_digest_matches(const char *rsc_id, const char *agent,
const char *digest_calc, const char *node_summary)
{
bool matches = FALSE;
if (rsc_id && agent && digest_calc && node_summary) {
char *search_secure = create_unfencing_summary(rsc_id, agent,
digest_calc);
/* The digest was calculated including the device ID and agent,
* so there is no risk of collision using strstr().
*/
matches = (strstr(node_summary, search_secure) != NULL);
crm_trace("Calculated unfencing digest '%s' %sfound in '%s'",
search_secure, matches? "" : "not ", node_summary);
free(search_secure);
}
return matches;
}
/* Magic string to use as action name for digest cache entries used for
* unfencing checks. This is not a real action name (i.e. "on"), so
* check_action_definition() won't confuse these entries with real actions.
*/
#define STONITH_DIGEST_TASK "stonith-on"
/*!
* \internal
* \brief Calculate fence device digests and digest comparison result
*
* \param[in] rsc Fence device resource
* \param[in] agent Fence device's agent type
* \param[in] node Node with digest cache to use
* \param[in] data_set Cluster working set
*
* \return Node's digest cache entry
*/
static op_digest_cache_t *
fencing_action_digest_cmp(pe_resource_t *rsc, const char *agent,
pe_node_t *node, pe_working_set_t *data_set)
{
const char *node_summary = NULL;
// Calculate device's current parameter digests
char *key = pcmk__op_key(rsc->id, STONITH_DIGEST_TASK, 0);
op_digest_cache_t *data = rsc_action_digest(rsc, STONITH_DIGEST_TASK, key,
node, NULL, TRUE, data_set);
free(key);
// Check whether node has special unfencing summary node attribute
node_summary = pe_node_attribute_raw(node, CRM_ATTR_DIGESTS_ALL);
if (node_summary == NULL) {
data->rc = RSC_DIGEST_UNKNOWN;
return data;
}
// Check whether full parameter digest matches
if (unfencing_digest_matches(rsc->id, agent, data->digest_all_calc,
node_summary)) {
data->rc = RSC_DIGEST_MATCH;
return data;
}
// Check whether secure parameter digest matches
node_summary = pe_node_attribute_raw(node, CRM_ATTR_DIGESTS_SECURE);
if (unfencing_digest_matches(rsc->id, agent, data->digest_secure_calc,
node_summary)) {
data->rc = RSC_DIGEST_MATCH;
if (is_set(data_set->flags, pe_flag_stdout)) {
printf("Only 'private' parameters to %s for unfencing %s changed\n",
rsc->id, node->details->uname);
}
return data;
}
// Parameters don't match
data->rc = RSC_DIGEST_ALL;
if (is_set(data_set->flags, (pe_flag_sanitized|pe_flag_stdout))
&& data->digest_secure_calc) {
char *digest = create_unfencing_summary(rsc->id, agent,
data->digest_secure_calc);
printf("Parameters to %s for unfencing %s changed, try '%s'\n",
rsc->id, node->details->uname, digest);
free(digest);
}
return data;
}
const char *rsc_printable_id(pe_resource_t *rsc)
{
if (is_not_set(rsc->flags, pe_rsc_unique)) {
return ID(rsc->xml);
}
return rsc->id;
}
void
clear_bit_recursive(pe_resource_t * rsc, unsigned long long flag)
{
GListPtr gIter = rsc->children;
clear_bit(rsc->flags, flag);
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
clear_bit_recursive(child_rsc, flag);
}
}
void
set_bit_recursive(pe_resource_t * rsc, unsigned long long flag)
{
GListPtr gIter = rsc->children;
set_bit(rsc->flags, flag);
for (; gIter != NULL; gIter = gIter->next) {
pe_resource_t *child_rsc = (pe_resource_t *) gIter->data;
set_bit_recursive(child_rsc, flag);
}
}
static GListPtr
find_unfencing_devices(GListPtr candidates, GListPtr matches)
{
for (GListPtr gIter = candidates; gIter != NULL; gIter = gIter->next) {
pe_resource_t *candidate = gIter->data;
const char *provides = g_hash_table_lookup(candidate->meta, XML_RSC_ATTR_PROVIDES);
const char *requires = g_hash_table_lookup(candidate->meta, XML_RSC_ATTR_REQUIRES);
if(candidate->children) {
matches = find_unfencing_devices(candidate->children, matches);
} else if (is_not_set(candidate->flags, pe_rsc_fence_device)) {
continue;
} else if (crm_str_eq(provides, "unfencing", FALSE) || crm_str_eq(requires, "unfencing", FALSE)) {
matches = g_list_prepend(matches, candidate);
}
}
return matches;
}
static int
node_priority_fencing_delay(pe_node_t * node, pe_working_set_t * data_set)
{
int member_count = 0;
int online_count = 0;
int top_priority = 0;
int lowest_priority = 0;
GListPtr gIter = NULL;
// `priority-fencing-delay` is disabled
if (data_set->priority_fencing_delay <= 0) {
return 0;
}
/* No need to request a delay if the fencing target is not a normal cluster
* member, for example if it's a remote node or a guest node. */
if (node->details->type != node_member) {
return 0;
}
// No need to request a delay if the fencing target is in our partition
if (node->details->online) {
return 0;
}
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *n = gIter->data;
if (n->details->type != node_member) {
continue;
}
member_count ++;
if (n->details->online) {
online_count++;
}
if (member_count == 1
|| n->details->priority > top_priority) {
top_priority = n->details->priority;
}
if (member_count == 1
|| n->details->priority < lowest_priority) {
lowest_priority = n->details->priority;
}
}
// No need to delay if we have more than half of the cluster members
if (online_count > member_count / 2) {
return 0;
}
/* All the nodes have equal priority.
* Any configured corresponding `pcmk_delay_base/max` will be applied. */
if (lowest_priority == top_priority) {
return 0;
}
if (node->details->priority < top_priority) {
return 0;
}
return data_set->priority_fencing_delay;
}
pe_action_t *
pe_fence_op(pe_node_t * node, const char *op, bool optional, const char *reason,
bool priority_delay, pe_working_set_t * data_set)
{
char *op_key = NULL;
pe_action_t *stonith_op = NULL;
if(op == NULL) {
op = data_set->stonith_action;
}
op_key = crm_strdup_printf("%s-%s-%s", CRM_OP_FENCE, node->details->uname, op);
if(data_set->singletons) {
stonith_op = g_hash_table_lookup(data_set->singletons, op_key);
}
if(stonith_op == NULL) {
stonith_op = custom_action(NULL, op_key, CRM_OP_FENCE, node, TRUE, TRUE, data_set);
add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET, node->details->uname);
add_hash_param(stonith_op->meta, XML_LRM_ATTR_TARGET_UUID, node->details->id);
add_hash_param(stonith_op->meta, "stonith_action", op);
if (pe__is_guest_or_remote_node(node)
&& is_set(data_set->flags, pe_flag_enable_unfencing)) {
/* Extra work to detect device changes on remotes
*
* We may do this for all nodes in the future, but for now
* the check_action_definition() based stuff works fine.
*/
long max = 1024;
long digests_all_offset = 0;
long digests_secure_offset = 0;
char *digests_all = calloc(max, sizeof(char));
char *digests_secure = calloc(max, sizeof(char));
GListPtr matches = find_unfencing_devices(data_set->resources, NULL);
for (GListPtr gIter = matches; gIter != NULL; gIter = gIter->next) {
pe_resource_t *match = gIter->data;
const char *agent = g_hash_table_lookup(match->meta,
XML_ATTR_TYPE);
op_digest_cache_t *data = NULL;
data = fencing_action_digest_cmp(match, agent, node, data_set);
if(data->rc == RSC_DIGEST_ALL) {
optional = FALSE;
crm_notice("Unfencing %s (remote): because the definition of %s changed", node->details->uname, match->id);
if (is_set(data_set->flags, pe_flag_stdout)) {
fprintf(stdout, " notice: Unfencing %s (remote): because the definition of %s changed\n", node->details->uname, match->id);
}
}
digests_all_offset += snprintf(
digests_all+digests_all_offset, max-digests_all_offset,
"%s:%s:%s,", match->id, agent, data->digest_all_calc);
digests_secure_offset += snprintf(
digests_secure+digests_secure_offset, max-digests_secure_offset,
"%s:%s:%s,", match->id, agent, data->digest_secure_calc);
}
g_hash_table_insert(stonith_op->meta,
strdup(XML_OP_ATTR_DIGESTS_ALL),
digests_all);
g_hash_table_insert(stonith_op->meta,
strdup(XML_OP_ATTR_DIGESTS_SECURE),
digests_secure);
}
} else {
free(op_key);
}
if (data_set->priority_fencing_delay > 0
/* It's a suitable case where `priority-fencing-delay` applies.
* At least add `priority-fencing-delay` field as an indicator. */
&& (priority_delay
/* Re-calculate priority delay for the suitable case when
* pe_fence_op() is called again by stage6() after node priority has
* been actually calculated with native_add_running() */
|| g_hash_table_lookup(stonith_op->meta,
XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY) != NULL)) {
/* Add `priority-fencing-delay` to the fencing op even if it's 0 for
* the targeting node. So that it takes precedence over any possible
* `pcmk_delay_base/max`.
*/
char *delay_s = crm_itoa(node_priority_fencing_delay(node, data_set));
g_hash_table_insert(stonith_op->meta,
strdup(XML_CONFIG_ATTR_PRIORITY_FENCING_DELAY),
delay_s);
}
if(optional == FALSE && pe_can_fence(data_set, node)) {
pe_action_required(stonith_op, NULL, reason);
} else if(reason && stonith_op->reason == NULL) {
stonith_op->reason = strdup(reason);
}
return stonith_op;
}
void
trigger_unfencing(
pe_resource_t * rsc, pe_node_t *node, const char *reason, pe_action_t *dependency, pe_working_set_t * data_set)
{
if(is_not_set(data_set->flags, pe_flag_enable_unfencing)) {
/* No resources require it */
return;
} else if (rsc != NULL && is_not_set(rsc->flags, pe_rsc_fence_device)) {
/* Wasn't a stonith device */
return;
} else if(node
&& node->details->online
&& node->details->unclean == FALSE
&& node->details->shutdown == FALSE) {
pe_action_t *unfence = pe_fence_op(node, "on", FALSE, reason, FALSE, data_set);
if(dependency) {
order_actions(unfence, dependency, pe_order_optional);
}
} else if(rsc) {
GHashTableIter iter;
g_hash_table_iter_init(&iter, rsc->allowed_nodes);
while (g_hash_table_iter_next(&iter, NULL, (void **)&node)) {
if(node->details->online && node->details->unclean == FALSE && node->details->shutdown == FALSE) {
trigger_unfencing(rsc, node, reason, dependency, data_set);
}
}
}
}
gboolean
add_tag_ref(GHashTable * tags, const char * tag_name, const char * obj_ref)
{
pe_tag_t *tag = NULL;
GListPtr gIter = NULL;
gboolean is_existing = FALSE;
CRM_CHECK(tags && tag_name && obj_ref, return FALSE);
tag = g_hash_table_lookup(tags, tag_name);
if (tag == NULL) {
tag = calloc(1, sizeof(pe_tag_t));
if (tag == NULL) {
return FALSE;
}
tag->id = strdup(tag_name);
tag->refs = NULL;
g_hash_table_insert(tags, strdup(tag_name), tag);
}
for (gIter = tag->refs; gIter != NULL; gIter = gIter->next) {
const char *existing_ref = (const char *) gIter->data;
if (crm_str_eq(existing_ref, obj_ref, TRUE)){
is_existing = TRUE;
break;
}
}
if (is_existing == FALSE) {
tag->refs = g_list_append(tag->refs, strdup(obj_ref));
crm_trace("Added: tag=%s ref=%s", tag->id, obj_ref);
}
return TRUE;
}
void pe_action_set_flag_reason(const char *function, long line,
pe_action_t *action, pe_action_t *reason, const char *text,
enum pe_action_flags flags, bool overwrite)
{
bool unset = FALSE;
bool update = FALSE;
const char *change = NULL;
if(is_set(flags, pe_action_runnable)) {
unset = TRUE;
change = "unrunnable";
} else if(is_set(flags, pe_action_optional)) {
unset = TRUE;
change = "required";
} else if(is_set(flags, pe_action_migrate_runnable)) {
unset = TRUE;
overwrite = TRUE;
change = "unrunnable";
} else if(is_set(flags, pe_action_dangle)) {
change = "dangling";
} else if(is_set(flags, pe_action_requires_any)) {
change = "required";
} else {
crm_err("Unknown flag change to %x by %s: 0x%s",
flags, action->uuid, (reason? reason->uuid : "0"));
}
if(unset) {
if(is_set(action->flags, flags)) {
action->flags = crm_clear_bit(function, line, action->uuid, action->flags, flags);
update = TRUE;
}
} else {
if(is_not_set(action->flags, flags)) {
action->flags = crm_set_bit(function, line, action->uuid, action->flags, flags);
update = TRUE;
}
}
if((change && update) || text) {
char *reason_text = NULL;
if(reason == NULL) {
pe_action_set_reason(action, text, overwrite);
} else if(reason->rsc == NULL) {
reason_text = crm_strdup_printf("%s %s%c %s", change, reason->task, text?':':0, text?text:"");
} else {
reason_text = crm_strdup_printf("%s %s %s%c %s", change, reason->rsc->id, reason->task, text?':':0, text?text:"NA");
}
if(reason_text && action->rsc != reason->rsc) {
pe_action_set_reason(action, reason_text, overwrite);
}
free(reason_text);
}
}
void pe_action_set_reason(pe_action_t *action, const char *reason, bool overwrite)
{
if (action->reason != NULL && overwrite) {
pe_rsc_trace(action->rsc, "Changing %s reason from '%s' to '%s'",
action->uuid, action->reason, crm_str(reason));
free(action->reason);
} else if (action->reason == NULL) {
pe_rsc_trace(action->rsc, "Set %s reason to '%s'",
action->uuid, crm_str(reason));
} else {
// crm_assert(action->reason != NULL && !overwrite);
return;
}
if (reason != NULL) {
action->reason = strdup(reason);
} else {
action->reason = NULL;
}
}
/*!
* \internal
* \brief Check whether shutdown has been requested for a node
*
* \param[in] node Node to check
*
* \return TRUE if node has shutdown attribute set and nonzero, FALSE otherwise
* \note This differs from simply using node->details->shutdown in that it can
* be used before that has been determined (and in fact to determine it),
* and it can also be used to distinguish requested shutdown from implicit
* shutdown of remote nodes by virtue of their connection stopping.
*/
bool
pe__shutdown_requested(pe_node_t *node)
{
const char *shutdown = pe_node_attribute_raw(node, XML_CIB_ATTR_SHUTDOWN);
return shutdown && strcmp(shutdown, "0");
}
/*!
* \internal
* \brief Update a data set's "recheck by" time
*
* \param[in] recheck Epoch time when recheck should happen
* \param[in,out] data_set Current working set
*/
void
pe__update_recheck_time(time_t recheck, pe_working_set_t *data_set)
{
if ((recheck > get_effective_time(data_set))
&& ((data_set->recheck_by == 0)
|| (data_set->recheck_by > recheck))) {
data_set->recheck_by = recheck;
}
}
/*!
* \internal
* \brief Wrapper for pe_unpack_nvpairs() using a cluster working set
*/
void
pe__unpack_dataset_nvpairs(xmlNode *xml_obj, const char *set_name,
pe_rule_eval_data_t *rule_data, GHashTable *hash,
const char *always_first, gboolean overwrite,
pe_working_set_t *data_set)
{
crm_time_t *next_change = crm_time_new_undefined();
pe_eval_nvpairs(data_set->input, xml_obj, set_name, rule_data, hash,
always_first, overwrite, next_change);
if (crm_time_is_defined(next_change)) {
time_t recheck = (time_t) crm_time_get_seconds_since_epoch(next_change);
pe__update_recheck_time(recheck, data_set);
}
crm_time_free(next_change);
}
bool
pe__resource_is_disabled(pe_resource_t *rsc)
{
const char *target_role = NULL;
CRM_CHECK(rsc != NULL, return false);
target_role = g_hash_table_lookup(rsc->meta, XML_RSC_ATTR_TARGET_ROLE);
if (target_role) {
enum rsc_role_e target_role_e = text2role(target_role);
if ((target_role_e == RSC_ROLE_STOPPED)
|| ((target_role_e == RSC_ROLE_SLAVE)
&& is_set(uber_parent(rsc)->flags, pe_rsc_promotable))) {
return true;
}
}
return false;
}
/*!
* \internal
* \brief Create an action to clear a resource's history from CIB
*
* \param[in] rsc Resource to clear
* \param[in] node Node to clear history on
*
* \return New action to clear resource history
*/
pe_action_t *
pe__clear_resource_history(pe_resource_t *rsc, pe_node_t *node,
pe_working_set_t *data_set)
{
char *key = NULL;
CRM_ASSERT(rsc && node);
key = pcmk__op_key(rsc->id, CRM_OP_LRM_DELETE, 0);
return custom_action(rsc, key, CRM_OP_LRM_DELETE, node, FALSE, TRUE,
data_set);
}
bool
pe__rsc_running_on_any_node_in_list(pe_resource_t *rsc, GListPtr node_list)
{
for (GListPtr ele = rsc->running_on; ele; ele = ele->next) {
pe_node_t *node = (pe_node_t *) ele->data;
if (pcmk__str_in_list(node_list, node->details->uname)) {
return true;
}
}
return false;
}
bool
pcmk__rsc_is_filtered(pe_resource_t *rsc, GListPtr only_show)
{
return (rsc->fns->active(rsc, FALSE) && !pe__rsc_running_on_any_node_in_list(rsc, only_show));
}
diff --git a/tools/crm_mon.c b/tools/crm_mon.c
index 6c50447000..b2e143b11f 100644
--- a/tools/crm_mon.c
+++ b/tools/crm_mon.c
@@ -1,2202 +1,2198 @@
/*
* Copyright 2004-2020 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 <crm/crm.h>
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
#include <stdlib.h>
#include <errno.h>
#include <fcntl.h>
#include <libgen.h>
#include <signal.h>
#include <sys/utsname.h>
#include <crm/msg_xml.h>
#include <crm/services.h>
#include <crm/lrmd.h>
#include <crm/common/cmdline_internal.h>
#include <crm/common/curses_internal.h>
#include <crm/common/internal.h> // pcmk__ends_with_ext()
#include <crm/common/ipc.h>
#include <crm/common/iso8601_internal.h>
#include <crm/common/mainloop.h>
#include <crm/common/output.h>
#include <crm/common/util.h>
#include <crm/common/xml.h>
#include <crm/cib/internal.h>
#include <crm/pengine/status.h>
#include <crm/pengine/internal.h>
#include <pacemaker-internal.h>
#include <crm/stonith-ng.h>
#include <crm/fencing/internal.h>
#include "crm_mon.h"
#define SUMMARY "Provides a summary of cluster's current state.\n\n" \
"Outputs varying levels of detail in a number of different formats."
/*
* Definitions indicating which items to print
*/
static unsigned int show;
/*
* Definitions indicating how to output
*/
static mon_output_format_t output_format = mon_output_unset;
/* other globals */
static GIOChannel *io_channel = NULL;
static GMainLoop *mainloop = NULL;
static guint timer_id = 0;
static mainloop_timer_t *refresh_timer = NULL;
static pe_working_set_t *mon_data_set = NULL;
static cib_t *cib = NULL;
static stonith_t *st = NULL;
static xmlNode *current_cib = NULL;
static GError *error = NULL;
static pcmk__common_args_t *args = NULL;
static pcmk__output_t *out = NULL;
static GOptionContext *context = NULL;
static gchar **processed_args = NULL;
static time_t last_refresh = 0;
crm_trigger_t *refresh_trigger = NULL;
static pcmk__supported_format_t formats[] = {
#if CURSES_ENABLED
CRM_MON_SUPPORTED_FORMAT_CURSES,
#endif
PCMK__SUPPORTED_FORMAT_HTML,
PCMK__SUPPORTED_FORMAT_NONE,
PCMK__SUPPORTED_FORMAT_TEXT,
CRM_MON_SUPPORTED_FORMAT_XML,
{ NULL, NULL, NULL }
};
/* Define exit codes for monitoring-compatible output
* For nagios plugins, the possibilities are
* OK=0, WARN=1, CRIT=2, and UNKNOWN=3
*/
#define MON_STATUS_WARN CRM_EX_ERROR
#define MON_STATUS_CRIT CRM_EX_INVALID_PARAM
#define MON_STATUS_UNKNOWN CRM_EX_UNIMPLEMENT_FEATURE
#define RECONNECT_MSECS 5000
struct {
int reconnect_msec;
gboolean daemonize;
gboolean show_bans;
char *pid_file;
char *external_agent;
char *external_recipient;
char *neg_location_prefix;
char *only_node;
unsigned int mon_ops;
GSList *user_includes_excludes;
GSList *includes_excludes;
} options = {
.reconnect_msec = RECONNECT_MSECS,
.mon_ops = mon_op_default
};
static void clean_up_connections(void);
static crm_exit_t clean_up(crm_exit_t exit_code);
static void crm_diff_update(const char *event, xmlNode * msg);
static gboolean mon_refresh_display(gpointer user_data);
static int cib_connect(gboolean full);
static void mon_st_callback_event(stonith_t * st, stonith_event_t * e);
static void mon_st_callback_display(stonith_t * st, stonith_event_t * e);
static void kick_refresh(gboolean data_updated);
static unsigned int
all_includes(mon_output_format_t fmt) {
if (fmt == mon_output_monitor || fmt == mon_output_plain || fmt == mon_output_console) {
return ~mon_show_options;
} else {
return mon_show_all;
}
}
static unsigned int
default_includes(mon_output_format_t fmt) {
switch (fmt) {
case mon_output_monitor:
case mon_output_plain:
case mon_output_console:
return mon_show_stack | mon_show_dc | mon_show_times | mon_show_counts |
mon_show_nodes | mon_show_resources | mon_show_failures;
- break;
case mon_output_xml:
case mon_output_legacy_xml:
return all_includes(fmt);
- break;
case mon_output_html:
case mon_output_cgi:
return mon_show_summary | mon_show_nodes | mon_show_resources |
mon_show_failures;
- break;
default:
return 0;
- break;
}
}
struct {
const char *name;
unsigned int bit;
} sections[] = {
{ "attributes", mon_show_attributes },
{ "bans", mon_show_bans },
{ "counts", mon_show_counts },
{ "dc", mon_show_dc },
{ "failcounts", mon_show_failcounts },
{ "failures", mon_show_failures },
{ "fencing", mon_show_fencing_all },
{ "fencing-failed", mon_show_fence_failed },
{ "fencing-pending", mon_show_fence_pending },
{ "fencing-succeeded", mon_show_fence_worked },
{ "nodes", mon_show_nodes },
{ "operations", mon_show_operations },
{ "options", mon_show_options },
{ "resources", mon_show_resources },
{ "stack", mon_show_stack },
{ "summary", mon_show_summary },
{ "tickets", mon_show_tickets },
{ "times", mon_show_times },
{ NULL }
};
static unsigned int
find_section_bit(const char *name) {
for (int i = 0; sections[i].name != NULL; i++) {
if (crm_str_eq(sections[i].name, name, FALSE)) {
return sections[i].bit;
}
}
return 0;
}
static gboolean
apply_exclude(const gchar *excludes, GError **error) {
char **parts = NULL;
parts = g_strsplit(excludes, ",", 0);
for (char **s = parts; *s != NULL; s++) {
unsigned int bit = find_section_bit(*s);
if (crm_str_eq(*s, "all", TRUE)) {
show = 0;
} else if (crm_str_eq(*s, "none", TRUE)) {
show = all_includes(output_format);
} else if (bit != 0) {
show &= ~bit;
} else {
g_set_error(error, G_OPTION_ERROR, CRM_EX_USAGE,
"--exclude options: all, attributes, bans, counts, dc, "
"failcounts, failures, fencing, fencing-failed, "
"fencing-pending, fencing-succeeded, nodes, none, "
"operations, options, resources, stack, summary, "
"tickets, times");
return FALSE;
}
}
g_strfreev(parts);
return TRUE;
}
static gboolean
apply_include(const gchar *includes, GError **error) {
char **parts = NULL;
parts = g_strsplit(includes, ",", 0);
for (char **s = parts; *s != NULL; s++) {
unsigned int bit = find_section_bit(*s);
if (crm_str_eq(*s, "all", TRUE)) {
show = all_includes(output_format);
} else if (pcmk__starts_with(*s, "bans")) {
show |= mon_show_bans;
if (options.neg_location_prefix != NULL) {
free(options.neg_location_prefix);
options.neg_location_prefix = NULL;
}
if (strlen(*s) > 4 && (*s)[4] == ':') {
options.neg_location_prefix = strdup(*s+5);
}
} else if (crm_str_eq(*s, "default", TRUE) || crm_str_eq(*s, "defaults", TRUE)) {
show |= default_includes(output_format);
} else if (crm_str_eq(*s, "none", TRUE)) {
show = 0;
} else if (bit != 0) {
show |= bit;
} else {
g_set_error(error, G_OPTION_ERROR, CRM_EX_USAGE,
"--include options: all, attributes, bans[:PREFIX], counts, dc, "
"default, failcounts, failures, fencing, fencing-failed, "
"fencing-pending, fencing-succeeded, nodes, none, operations, "
"options, resources, stack, summary, tickets, times");
return FALSE;
}
}
g_strfreev(parts);
return TRUE;
}
static gboolean
apply_include_exclude(GSList *lst, mon_output_format_t fmt, GError **error) {
gboolean rc = TRUE;
GSList *node = lst;
/* Set the default of what to display here. Note that we OR everything to
* show instead of set show directly because it could have already had some
* settings applied to it in main.
*/
show |= default_includes(fmt);
while (node != NULL) {
char *s = node->data;
if (pcmk__starts_with(s, "--include=")) {
rc = apply_include(s+10, error);
} else if (pcmk__starts_with(s, "-I=")) {
rc = apply_include(s+3, error);
} else if (pcmk__starts_with(s, "--exclude=")) {
rc = apply_exclude(s+10, error);
} else if (pcmk__starts_with(s, "-U=")) {
rc = apply_exclude(s+3, error);
}
if (rc != TRUE) {
break;
}
node = node->next;
}
return rc;
}
static gboolean
user_include_exclude_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
char *s = crm_strdup_printf("%s=%s", option_name, optarg);
options.user_includes_excludes = g_slist_append(options.user_includes_excludes, s);
return TRUE;
}
static gboolean
include_exclude_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
char *s = crm_strdup_printf("%s=%s", option_name, optarg);
options.includes_excludes = g_slist_append(options.includes_excludes, s);
return TRUE;
}
static gboolean
as_cgi_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
if (args->output_ty != NULL) {
free(args->output_ty);
}
args->output_ty = strdup("html");
output_format = mon_output_cgi;
options.mon_ops |= mon_op_one_shot;
return TRUE;
}
static gboolean
as_html_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
if (args->output_ty != NULL) {
free(args->output_ty);
}
if (args->output_dest != NULL) {
free(args->output_dest);
args->output_dest = NULL;
}
if (optarg != NULL) {
args->output_dest = strdup(optarg);
}
args->output_ty = strdup("html");
output_format = mon_output_html;
umask(S_IWGRP | S_IWOTH);
return TRUE;
}
static gboolean
as_simple_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
if (args->output_ty != NULL) {
free(args->output_ty);
}
args->output_ty = strdup("text");
output_format = mon_output_monitor;
options.mon_ops |= mon_op_one_shot;
return TRUE;
}
static gboolean
as_xml_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
if (args->output_ty != NULL) {
free(args->output_ty);
}
args->output_ty = strdup("xml");
output_format = mon_output_legacy_xml;
options.mon_ops |= mon_op_one_shot;
return TRUE;
}
static gboolean
fence_history_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
int rc = crm_atoi(optarg, "2");
switch (rc) {
case 3:
options.mon_ops |= mon_op_fence_full_history | mon_op_fence_history | mon_op_fence_connect;
return include_exclude_cb("--include", "fencing", data, err);
case 2:
options.mon_ops |= mon_op_fence_history | mon_op_fence_connect;
return include_exclude_cb("--include", "fencing", data, err);
case 1:
options.mon_ops |= mon_op_fence_history | mon_op_fence_connect;
return include_exclude_cb("--include", "fencing-failed,fencing-pending", data, err);
case 0:
options.mon_ops &= ~(mon_op_fence_history | mon_op_fence_connect);
return include_exclude_cb("--exclude", "fencing", data, err);
default:
g_set_error(err, G_OPTION_ERROR, CRM_EX_INVALID_PARAM, "Fence history must be 0-3");
return FALSE;
}
}
static gboolean
group_by_node_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
options.mon_ops |= mon_op_group_by_node;
return TRUE;
}
static gboolean
hide_headers_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
return include_exclude_cb("--exclude", "summary", data, err);
}
static gboolean
inactive_resources_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
options.mon_ops |= mon_op_inactive_resources;
return TRUE;
}
static gboolean
no_curses_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
output_format = mon_output_plain;
return TRUE;
}
static gboolean
one_shot_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
options.mon_ops |= mon_op_one_shot;
return TRUE;
}
static gboolean
print_brief_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
options.mon_ops |= mon_op_print_brief;
return TRUE;
}
static gboolean
print_clone_detail_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
options.mon_ops |= mon_op_print_clone_detail;
return TRUE;
}
static gboolean
print_pending_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
options.mon_ops |= mon_op_print_pending;
return TRUE;
}
static gboolean
print_timing_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
options.mon_ops |= mon_op_print_timing;
return include_exclude_cb("--include", "operations", data, err);
}
static gboolean
reconnect_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
int rc = crm_get_msec(optarg);
if (rc == -1) {
g_set_error(err, G_OPTION_ERROR, CRM_EX_INVALID_PARAM, "Invalid value for -i: %s", optarg);
return FALSE;
} else {
options.reconnect_msec = crm_parse_interval_spec(optarg);
}
return TRUE;
}
static gboolean
show_attributes_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
return include_exclude_cb("--include", "attributes", data, err);
}
static gboolean
show_bans_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
if (optarg != NULL) {
char *s = crm_strdup_printf("bans:%s", optarg);
gboolean rc = include_exclude_cb("--include", s, data, err);
free(s);
return rc;
} else {
return include_exclude_cb("--include", "bans", data, err);
}
}
static gboolean
show_failcounts_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
return include_exclude_cb("--include", "failcounts", data, err);
}
static gboolean
show_operations_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
return include_exclude_cb("--include", "failcounts,operations", data, err);
}
static gboolean
show_tickets_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
return include_exclude_cb("--include", "tickets", data, err);
}
static gboolean
use_cib_file_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
setenv("CIB_file", optarg, 1);
options.mon_ops |= mon_op_one_shot;
return TRUE;
}
static gboolean
watch_fencing_cb(const gchar *option_name, const gchar *optarg, gpointer data, GError **err) {
options.mon_ops |= mon_op_watch_fencing;
return TRUE;
}
#define INDENT " "
/* *INDENT-OFF* */
static GOptionEntry addl_entries[] = {
{ "interval", 'i', 0, G_OPTION_ARG_CALLBACK, reconnect_cb,
"Update frequency (default is 5 seconds)",
"TIMESPEC" },
{ "one-shot", '1', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, one_shot_cb,
"Display the cluster status once on the console and exit",
NULL },
{ "daemonize", 'd', 0, G_OPTION_ARG_NONE, &options.daemonize,
"Run in the background as a daemon.\n"
INDENT "Requires at least one of --output-to and --external-agent.",
NULL },
{ "pid-file", 'p', 0, G_OPTION_ARG_FILENAME, &options.pid_file,
"(Advanced) Daemon pid file location",
"FILE" },
{ "external-agent", 'E', 0, G_OPTION_ARG_FILENAME, &options.external_agent,
"A program to run when resource operations take place",
"FILE" },
{ "external-recipient", 'e', 0, G_OPTION_ARG_STRING, &options.external_recipient,
"A recipient for your program (assuming you want the program to send something to someone).",
"RCPT" },
{ "watch-fencing", 'W', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, watch_fencing_cb,
"Listen for fencing events. For use with --external-agent.",
NULL },
{ "xml-file", 'x', G_OPTION_FLAG_HIDDEN, G_OPTION_ARG_CALLBACK, use_cib_file_cb,
NULL,
NULL },
{ NULL }
};
static GOptionEntry display_entries[] = {
{ "include", 'I', 0, G_OPTION_ARG_CALLBACK, user_include_exclude_cb,
"A list of sections to include in the output.\n"
INDENT "See `Output Control` help for more information.",
"SECTION(s)" },
{ "exclude", 'U', 0, G_OPTION_ARG_CALLBACK, user_include_exclude_cb,
"A list of sections to exclude from the output.\n"
INDENT "See `Output Control` help for more information.",
"SECTION(s)" },
{ "node", 0, 0, G_OPTION_ARG_STRING, &options.only_node,
"When displaying information about nodes, show only what's related to the given\n"
INDENT "node, or to all nodes tagged with the given tag",
"NODE" },
{ "group-by-node", 'n', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, group_by_node_cb,
"Group resources by node",
NULL },
{ "inactive", 'r', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, inactive_resources_cb,
"Display inactive resources",
NULL },
{ "failcounts", 'f', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_failcounts_cb,
"Display resource fail counts",
NULL },
{ "operations", 'o', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_operations_cb,
"Display resource operation history",
NULL },
{ "timing-details", 't', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_timing_cb,
"Display resource operation history with timing details",
NULL },
{ "tickets", 'c', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_tickets_cb,
"Display cluster tickets",
NULL },
{ "fence-history", 'm', G_OPTION_FLAG_OPTIONAL_ARG, G_OPTION_ARG_CALLBACK, fence_history_cb,
"Show fence history:\n"
INDENT "0=off, 1=failures and pending (default without option),\n"
INDENT "2=add successes (default without value for option),\n"
INDENT "3=show full history without reduction to most recent of each flavor",
"LEVEL" },
{ "neg-locations", 'L', G_OPTION_FLAG_OPTIONAL_ARG, G_OPTION_ARG_CALLBACK, show_bans_cb,
"Display negative location constraints [optionally filtered by id prefix]",
NULL },
{ "show-node-attributes", 'A', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, show_attributes_cb,
"Display node attributes",
NULL },
{ "hide-headers", 'D', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, hide_headers_cb,
"Hide all headers",
NULL },
{ "show-detail", 'R', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_clone_detail_cb,
"Show more details (node IDs, individual clone instances)",
NULL },
{ "brief", 'b', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_brief_cb,
"Brief output",
NULL },
{ "pending", 'j', G_OPTION_FLAG_HIDDEN|G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, print_pending_cb,
"Display pending state if 'record-pending' is enabled",
NULL },
{ "simple-status", 's', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, as_simple_cb,
"Display the cluster status once as a simple one line output (suitable for nagios)",
NULL },
{ NULL }
};
static GOptionEntry deprecated_entries[] = {
{ "as-html", 'h', G_OPTION_FLAG_FILENAME, G_OPTION_ARG_CALLBACK, as_html_cb,
"Write cluster status to the named HTML file.\n"
INDENT "Use --output-as=html --output-to=FILE instead.",
"FILE" },
{ "as-xml", 'X', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, as_xml_cb,
"Write cluster status as XML to stdout. This will enable one-shot mode.\n"
INDENT "Use --output-as=xml instead.",
NULL },
{ "disable-ncurses", 'N', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, no_curses_cb,
"Disable the use of ncurses.\n"
INDENT "Use --output-as=text instead.",
NULL },
{ "web-cgi", 'w', G_OPTION_FLAG_NO_ARG, G_OPTION_ARG_CALLBACK, as_cgi_cb,
"Web mode with output suitable for CGI (preselected when run as *.cgi).\n"
INDENT "Use --output-as=html --html-cgi instead.",
NULL },
{ NULL }
};
/* *INDENT-ON* */
static gboolean
mon_timer_popped(gpointer data)
{
int rc = pcmk_ok;
#if CURSES_ENABLED
if (output_format == mon_output_console) {
clear();
refresh();
}
#endif
if (timer_id > 0) {
g_source_remove(timer_id);
timer_id = 0;
}
print_as(output_format, "Reconnecting...\n");
rc = cib_connect(TRUE);
if (rc != pcmk_ok) {
timer_id = g_timeout_add(options.reconnect_msec, mon_timer_popped, NULL);
}
return FALSE;
}
static void
do_mon_cib_connection_destroy(gpointer user_data, bool is_error)
{
if (is_error) {
out->err(out, "Connection to the cluster-daemons terminated");
} else {
out->info(out, "Connection to the cluster-daemons terminated");
}
if (refresh_timer != NULL) {
/* we'll trigger a refresh after reconnect */
mainloop_timer_stop(refresh_timer);
}
if (timer_id) {
/* we'll trigger a new reconnect-timeout at the end */
g_source_remove(timer_id);
timer_id = 0;
}
if (st) {
/* the client API won't properly reconnect notifications
* if they are still in the table - so remove them
*/
st->cmds->remove_notification(st, T_STONITH_NOTIFY_DISCONNECT);
st->cmds->remove_notification(st, T_STONITH_NOTIFY_FENCE);
st->cmds->remove_notification(st, T_STONITH_NOTIFY_HISTORY);
if (st->state != stonith_disconnected) {
st->cmds->disconnect(st);
}
}
if (cib) {
cib->cmds->signoff(cib);
timer_id = g_timeout_add(options.reconnect_msec, mon_timer_popped, NULL);
}
return;
}
static void
mon_cib_connection_destroy_regular(gpointer user_data)
{
do_mon_cib_connection_destroy(user_data, false);
}
static void
mon_cib_connection_destroy_error(gpointer user_data)
{
do_mon_cib_connection_destroy(user_data, true);
}
/*
* Mainloop signal handler.
*/
static void
mon_shutdown(int nsig)
{
clean_up(CRM_EX_OK);
}
#if CURSES_ENABLED
static sighandler_t ncurses_winch_handler;
static void
mon_winresize(int nsig)
{
static int not_done;
int lines = 0, cols = 0;
if (!not_done++) {
if (ncurses_winch_handler)
/* the original ncurses WINCH signal handler does the
* magic of retrieving the new window size;
* otherwise, we'd have to use ioctl or tgetent */
(*ncurses_winch_handler) (SIGWINCH);
getmaxyx(stdscr, lines, cols);
resizeterm(lines, cols);
mainloop_set_trigger(refresh_trigger);
}
not_done--;
}
#endif
static int
cib_connect(gboolean full)
{
int rc = pcmk_ok;
static gboolean need_pass = TRUE;
CRM_CHECK(cib != NULL, return -EINVAL);
if (getenv("CIB_passwd") != NULL) {
need_pass = FALSE;
}
if (is_set(options.mon_ops, mon_op_fence_connect) && st == NULL) {
st = stonith_api_new();
}
if (is_set(options.mon_ops, mon_op_fence_connect) && st != NULL && st->state == stonith_disconnected) {
rc = st->cmds->connect(st, crm_system_name, NULL);
if (rc == pcmk_ok) {
crm_trace("Setting up stonith callbacks");
if (is_set(options.mon_ops, mon_op_watch_fencing)) {
st->cmds->register_notification(st, T_STONITH_NOTIFY_DISCONNECT,
mon_st_callback_event);
st->cmds->register_notification(st, T_STONITH_NOTIFY_FENCE, mon_st_callback_event);
} else {
st->cmds->register_notification(st, T_STONITH_NOTIFY_DISCONNECT,
mon_st_callback_display);
st->cmds->register_notification(st, T_STONITH_NOTIFY_HISTORY, mon_st_callback_display);
}
}
}
if (cib->state != cib_connected_query && cib->state != cib_connected_command) {
crm_trace("Connecting to the CIB");
/* Hack: the CIB signon will print the prompt for a password if needed,
* but to stderr. If we're in curses, show it on the screen instead.
*
* @TODO Add a password prompt (maybe including input) function to
* pcmk__output_t and use it in libcib.
*/
if ((output_format == mon_output_console) && need_pass && (cib->variant == cib_remote)) {
need_pass = FALSE;
print_as(output_format, "Password:");
}
rc = cib->cmds->signon(cib, crm_system_name, cib_query);
if (rc != pcmk_ok) {
out->err(out, "Could not connect to the CIB: %s",
pcmk_strerror(rc));
return rc;
}
rc = cib->cmds->query(cib, NULL, &current_cib, cib_scope_local | cib_sync_call);
if (rc == pcmk_ok) {
mon_refresh_display(&output_format);
}
if (rc == pcmk_ok && full) {
if (rc == pcmk_ok) {
rc = cib->cmds->set_connection_dnotify(cib, mon_cib_connection_destroy_regular);
if (rc == -EPROTONOSUPPORT) {
print_as
(output_format, "Notification setup not supported, won't be able to reconnect after failure");
if (output_format == mon_output_console) {
sleep(2);
}
rc = pcmk_ok;
}
}
if (rc == pcmk_ok) {
cib->cmds->del_notify_callback(cib, T_CIB_DIFF_NOTIFY, crm_diff_update);
rc = cib->cmds->add_notify_callback(cib, T_CIB_DIFF_NOTIFY, crm_diff_update);
}
if (rc != pcmk_ok) {
out->err(out, "Notification setup failed, could not monitor CIB actions");
clean_up_connections();
}
}
}
return rc;
}
#if CURSES_ENABLED
static const char *
get_option_desc(char c)
{
const char *desc = "No help available";
for (GOptionEntry *entry = display_entries; entry != NULL; entry++) {
if (entry->short_name == c) {
desc = entry->description;
break;
}
}
return desc;
}
#define print_option_help(output_format, option, condition) \
out->info(out, "%c %c: \t%s", ((condition)? '*': ' '), option, get_option_desc(option));
static gboolean
detect_user_input(GIOChannel *channel, GIOCondition condition, gpointer user_data)
{
int c;
gboolean config_mode = FALSE;
while (1) {
/* Get user input */
c = getchar();
switch (c) {
case 'm':
if (is_not_set(show, mon_show_fencing_all)) {
options.mon_ops |= mon_op_fence_history;
options.mon_ops |= mon_op_fence_connect;
if (st == NULL) {
mon_cib_connection_destroy_regular(NULL);
}
}
if (is_set(show, mon_show_fence_failed) || is_set(show, mon_show_fence_pending) ||
is_set(show, mon_show_fence_worked)) {
show &= ~mon_show_fencing_all;
} else {
show |= mon_show_fencing_all;
}
break;
case 'c':
show ^= mon_show_tickets;
break;
case 'f':
show ^= mon_show_failcounts;
break;
case 'n':
options.mon_ops ^= mon_op_group_by_node;
break;
case 'o':
show ^= mon_show_operations;
if (is_not_set(show, mon_show_operations)) {
options.mon_ops &= ~mon_op_print_timing;
}
break;
case 'r':
options.mon_ops ^= mon_op_inactive_resources;
break;
case 'R':
options.mon_ops ^= mon_op_print_clone_detail;
break;
case 't':
options.mon_ops ^= mon_op_print_timing;
if (is_set(options.mon_ops, mon_op_print_timing)) {
show |= mon_show_operations;
}
break;
case 'A':
show ^= mon_show_attributes;
break;
case 'L':
show ^= mon_show_bans;
break;
case 'D':
/* If any header is shown, clear them all, otherwise set them all */
if (is_set(show, mon_show_stack) || is_set(show, mon_show_dc) ||
is_set(show, mon_show_times) || is_set(show, mon_show_counts)) {
show &= ~mon_show_summary;
} else {
show |= mon_show_summary;
}
/* Regardless, we don't show options in console mode. */
show &= ~mon_show_options;
break;
case 'b':
options.mon_ops ^= mon_op_print_brief;
break;
case 'j':
options.mon_ops ^= mon_op_print_pending;
break;
case '?':
config_mode = TRUE;
break;
default:
goto refresh;
}
if (!config_mode)
goto refresh;
blank_screen();
out->info(out, "%s", "Display option change mode\n");
print_option_help(out, 'c', is_set(show, mon_show_tickets));
print_option_help(out, 'f', is_set(show, mon_show_failcounts));
print_option_help(out, 'n', is_set(options.mon_ops, mon_op_group_by_node));
print_option_help(out, 'o', is_set(show, mon_show_operations));
print_option_help(out, 'r', is_set(options.mon_ops, mon_op_inactive_resources));
print_option_help(out, 't', is_set(options.mon_ops, mon_op_print_timing));
print_option_help(out, 'A', is_set(show, mon_show_attributes));
print_option_help(out, 'L', is_set(show,mon_show_bans));
print_option_help(out, 'D', is_not_set(show, mon_show_summary));
print_option_help(out, 'R', is_set(options.mon_ops, mon_op_print_clone_detail));
print_option_help(out, 'b', is_set(options.mon_ops, mon_op_print_brief));
print_option_help(out, 'j', is_set(options.mon_ops, mon_op_print_pending));
print_option_help(out, 'm', is_set(show, mon_show_fencing_all));
out->info(out, "%s", "\nToggle fields via field letter, type any other key to return");
}
refresh:
mon_refresh_display(NULL);
return TRUE;
}
#endif
// Basically crm_signal_handler(SIGCHLD, SIG_IGN) plus the SA_NOCLDWAIT flag
static void
avoid_zombies(void)
{
struct sigaction sa;
memset(&sa, 0, sizeof(struct sigaction));
if (sigemptyset(&sa.sa_mask) < 0) {
crm_warn("Cannot avoid zombies: %s", pcmk_strerror(errno));
return;
}
sa.sa_handler = SIG_IGN;
sa.sa_flags = SA_RESTART|SA_NOCLDWAIT;
if (sigaction(SIGCHLD, &sa, NULL) < 0) {
crm_warn("Cannot avoid zombies: %s", pcmk_strerror(errno));
}
}
static GOptionContext *
build_arg_context(pcmk__common_args_t *args, GOptionGroup **group) {
GOptionContext *context = NULL;
GOptionEntry extra_prog_entries[] = {
{ "quiet", 'Q', 0, G_OPTION_ARG_NONE, &(args->quiet),
"Be less descriptive in output.",
NULL },
{ NULL }
};
const char *description = "Notes:\n\n"
"If this program is called as crm_mon.cgi, --output-as=html --html-cgi will\n"
"automatically be added to the command line arguments.\n\n"
"Time Specification:\n\n"
"The TIMESPEC in any command line option can be specified in many different\n"
"formats. It can be just an integer number of seconds, a number plus units\n"
"(ms/msec/us/usec/s/sec/m/min/h/hr), or an ISO 8601 period specification.\n\n"
"Output Control:\n\n"
"By default, a certain list of sections are written to the output destination.\n"
"The default varies based on the output format - XML includes everything, while\n"
"other output formats will display less. This list can be modified with the\n"
"--include and --exclude command line options. Each option may be given multiple\n"
"times on the command line, and each can give a comma-separated list of sections.\n"
"The options are applied to the default set, from left to right as seen on the\n"
"command line. For a list of valid sections, pass --include=list or --exclude=list.\n\n"
"Examples:\n\n"
"Display the cluster status on the console with updates as they occur:\n\n"
"\tcrm_mon\n\n"
"Display the cluster status on the console just once then exit:\n\n"
"\tcrm_mon -1\n\n"
"Display your cluster status, group resources by node, and include inactive resources in the list:\n\n"
"\tcrm_mon --group-by-node --inactive\n\n"
"Start crm_mon as a background daemon and have it write the cluster status to an HTML file:\n\n"
"\tcrm_mon --daemonize --output-as html --output-to /path/to/docroot/filename.html\n\n"
"Start crm_mon and export the current cluster status as XML to stdout, then exit:\n\n"
"\tcrm_mon --output-as xml\n\n";
context = pcmk__build_arg_context(args, "console (default), html, text, xml", group, NULL);
pcmk__add_main_args(context, extra_prog_entries);
g_option_context_set_description(context, description);
pcmk__add_arg_group(context, "display", "Display Options:",
"Show display options", display_entries);
pcmk__add_arg_group(context, "additional", "Additional Options:",
"Show additional options", addl_entries);
pcmk__add_arg_group(context, "deprecated", "Deprecated Options:",
"Show deprecated options", deprecated_entries);
return context;
}
/* If certain format options were specified, we want to set some extra
* options. We can just process these like they were given on the
* command line.
*/
static void
add_output_args(void) {
GError *err = NULL;
if (output_format == mon_output_plain) {
if (!pcmk__force_args(context, &err, "%s --text-fancy", g_get_prgname())) {
g_propagate_error(&error, err);
clean_up(CRM_EX_USAGE);
}
} else if (output_format == mon_output_cgi) {
if (!pcmk__force_args(context, &err, "%s --html-cgi", g_get_prgname())) {
g_propagate_error(&error, err);
clean_up(CRM_EX_USAGE);
}
} else if (output_format == mon_output_xml) {
if (!pcmk__force_args(context, &err, "%s --xml-simple-list", g_get_prgname())) {
g_propagate_error(&error, err);
clean_up(CRM_EX_USAGE);
}
} else if (output_format == mon_output_legacy_xml) {
output_format = mon_output_xml;
if (!pcmk__force_args(context, &err, "%s --xml-legacy", g_get_prgname())) {
g_propagate_error(&error, err);
clean_up(CRM_EX_USAGE);
}
}
}
/* Which output format to use could come from two places: The --as-xml
* style arguments we gave in deprecated_entries above, or the formatted output
* arguments added by pcmk__register_formats. If the latter were used,
* output_format will be mon_output_unset.
*
* Call the callbacks as if those older style arguments were provided so
* the various things they do get done.
*/
static void
reconcile_output_format(pcmk__common_args_t *args) {
gboolean retval = TRUE;
GError *err = NULL;
if (output_format != mon_output_unset) {
return;
}
if (safe_str_eq(args->output_ty, "html")) {
char *dest = NULL;
if (args->output_dest != NULL) {
dest = strdup(args->output_dest);
}
retval = as_html_cb("h", dest, NULL, &err);
free(dest);
} else if (safe_str_eq(args->output_ty, "text")) {
retval = no_curses_cb("N", NULL, NULL, &err);
} else if (safe_str_eq(args->output_ty, "xml")) {
if (args->output_ty != NULL) {
free(args->output_ty);
}
args->output_ty = strdup("xml");
output_format = mon_output_xml;
options.mon_ops |= mon_op_one_shot;
} else if (is_set(options.mon_ops, mon_op_one_shot)) {
if (args->output_ty != NULL) {
free(args->output_ty);
}
args->output_ty = strdup("text");
output_format = mon_output_plain;
} else {
/* Neither old nor new arguments were given, so set the default. */
if (args->output_ty != NULL) {
free(args->output_ty);
}
args->output_ty = strdup("console");
output_format = mon_output_console;
}
if (!retval) {
g_propagate_error(&error, err);
clean_up(CRM_EX_USAGE);
}
}
int
main(int argc, char **argv)
{
int rc = pcmk_ok;
GOptionGroup *output_group = NULL;
args = pcmk__new_common_args(SUMMARY);
context = build_arg_context(args, &output_group);
pcmk__register_formats(output_group, formats);
options.pid_file = strdup("/tmp/ClusterMon.pid");
crm_log_cli_init("crm_mon");
// Avoid needing to wait for subprocesses forked for -E/--external-agent
avoid_zombies();
if (pcmk__ends_with_ext(argv[0], ".cgi")) {
output_format = mon_output_cgi;
options.mon_ops |= mon_op_one_shot;
}
processed_args = pcmk__cmdline_preproc(argv, "ehimpxEILU");
fence_history_cb("--fence-history", "1", NULL, NULL);
/* Set an HTML title regardless of what format we will eventually use. This can't
* be done in add_output_args. That function is called after command line
* arguments are processed in the next block, which means it'll override whatever
* title the user provides. Doing this here means the user can give their own
* title on the command line.
*/
if (!pcmk__force_args(context, &error, "%s --html-title \"Cluster Status\"",
g_get_prgname())) {
return clean_up(CRM_EX_USAGE);
}
if (!g_option_context_parse_strv(context, &processed_args, &error)) {
return clean_up(CRM_EX_USAGE);
}
for (int i = 0; i < args->verbosity; i++) {
crm_bump_log_level(argc, argv);
}
if (!args->version) {
if (args->quiet) {
include_exclude_cb("--exclude", "times", NULL, NULL);
}
if (is_set(options.mon_ops, mon_op_watch_fencing)) {
fence_history_cb("--fence-history", "0", NULL, NULL);
options.mon_ops |= mon_op_fence_connect;
}
/* create the cib-object early to be able to do further
* decisions based on the cib-source
*/
cib = cib_new();
if (cib == NULL) {
rc = -EINVAL;
} else {
switch (cib->variant) {
case cib_native:
/* cib & fencing - everything available */
break;
case cib_file:
/* Don't try to connect to fencing as we
* either don't have a running cluster or
* the fencing-information would possibly
* not match the cib data from a file.
* As we don't expect cib-updates coming
* in enforce one-shot. */
fence_history_cb("--fence-history", "0", NULL, NULL);
options.mon_ops |= mon_op_one_shot;
break;
case cib_remote:
/* updates coming in but no fencing */
fence_history_cb("--fence-history", "0", NULL, NULL);
break;
case cib_undefined:
case cib_database:
default:
/* something is odd */
rc = -EINVAL;
break;
}
}
if (is_set(options.mon_ops, mon_op_one_shot)) {
if (output_format == mon_output_console) {
output_format = mon_output_plain;
}
} else if (options.daemonize) {
if ((output_format == mon_output_console) || (output_format == mon_output_plain)) {
output_format = mon_output_none;
}
crm_enable_stderr(FALSE);
if ((args->output_dest == NULL || safe_str_eq(args->output_dest, "-")) && !options.external_agent) {
g_set_error(&error, G_OPTION_ERROR, CRM_EX_USAGE, "--daemonize requires at least one of --output-to and --external-agent");
return clean_up(CRM_EX_USAGE);
}
if (cib) {
/* to be on the safe side don't have cib-object around
* when we are forking
*/
cib_delete(cib);
cib = NULL;
crm_make_daemon(crm_system_name, TRUE, options.pid_file);
cib = cib_new();
if (cib == NULL) {
rc = -EINVAL;
}
/* otherwise assume we've got the same cib-object we've just destroyed
* in our parent
*/
}
} else if (output_format == mon_output_console) {
#if CURSES_ENABLED
crm_enable_stderr(FALSE);
#else
options.mon_ops |= mon_op_one_shot;
output_format = mon_output_plain;
printf("Defaulting to one-shot mode\n");
printf("You need to have curses available at compile time to enable console mode\n");
#endif
}
}
if (rc != pcmk_ok) {
// Shouldn't really be possible
g_set_error(&error, G_OPTION_ERROR, CRM_EX_ERROR, "Invalid CIB source");
return clean_up(CRM_EX_ERROR);
}
reconcile_output_format(args);
add_output_args();
if (args->version && output_format == mon_output_console) {
/* Use the text output format here if we are in curses mode but were given
* --version. Displaying version information uses printf, and then we
* immediately exit. We don't want to initialize curses for that.
*/
rc = pcmk__output_new(&out, "text", args->output_dest, argv);
} else {
rc = pcmk__output_new(&out, args->output_ty, args->output_dest, argv);
}
if (rc != pcmk_rc_ok) {
g_set_error(&error, G_OPTION_ERROR, CRM_EX_ERROR, "Error creating output format %s: %s",
args->output_ty, pcmk_rc_str(rc));
return clean_up(CRM_EX_ERROR);
}
/* output_format MUST NOT BE CHANGED AFTER THIS POINT. */
/* Apply --include/--exclude flags we used internally. There's no error reporting
* here because this would be a programming error.
*/
apply_include_exclude(options.includes_excludes, output_format, &error);
/* And now apply any --include/--exclude flags the user gave on the command line.
* These are done in a separate pass from the internal ones because we want to
* make sure whatever the user specifies overrides whatever we do.
*/
if (!apply_include_exclude(options.user_includes_excludes, output_format, &error)) {
return clean_up(CRM_EX_USAGE);
}
crm_mon_register_messages(out);
pe__register_messages(out);
stonith__register_messages(out);
if (args->version) {
out->version(out, false);
return clean_up(CRM_EX_OK);
}
/* Extra sanity checks when in CGI mode */
if (output_format == mon_output_cgi) {
if (cib && cib->variant == cib_file) {
g_set_error(&error, G_OPTION_ERROR, CRM_EX_USAGE, "CGI mode used with CIB file");
return clean_up(CRM_EX_USAGE);
} else if (options.external_agent != NULL) {
g_set_error(&error, G_OPTION_ERROR, CRM_EX_USAGE, "CGI mode cannot be used with --external-agent");
return clean_up(CRM_EX_USAGE);
} else if (options.daemonize == TRUE) {
g_set_error(&error, G_OPTION_ERROR, CRM_EX_USAGE, "CGI mode cannot be used with -d");
return clean_up(CRM_EX_USAGE);
}
}
if (output_format == mon_output_xml || output_format == mon_output_legacy_xml) {
options.mon_ops |= mon_op_print_timing | mon_op_inactive_resources;
}
crm_info("Starting %s", crm_system_name);
if (cib) {
do {
if (is_not_set(options.mon_ops, mon_op_one_shot)) {
print_as(output_format ,"Waiting until cluster is available on this node ...\n");
}
rc = cib_connect(is_not_set(options.mon_ops, mon_op_one_shot));
if (is_set(options.mon_ops, mon_op_one_shot)) {
break;
} else if (rc != pcmk_ok) {
sleep(options.reconnect_msec / 1000);
#if CURSES_ENABLED
if (output_format == mon_output_console) {
clear();
refresh();
}
#endif
} else {
if (output_format == mon_output_html && out->dest != stdout) {
printf("Writing html to %s ...\n", args->output_dest);
}
}
} while (rc == -ENOTCONN);
}
if (rc != pcmk_ok) {
if (output_format == mon_output_monitor) {
g_set_error(&error, G_OPTION_ERROR, CRM_EX_ERROR, "CLUSTER CRIT: Connection to cluster failed: %s",
pcmk_strerror(rc));
return clean_up(MON_STATUS_CRIT);
} else {
if (rc == -ENOTCONN) {
g_set_error(&error, G_OPTION_ERROR, CRM_EX_ERROR, "Error: cluster is not available on this node");
} else {
g_set_error(&error, G_OPTION_ERROR, CRM_EX_ERROR, "Connection to cluster failed: %s", pcmk_strerror(rc));
}
}
return clean_up(crm_errno2exit(rc));
}
if (is_set(options.mon_ops, mon_op_one_shot)) {
return clean_up(CRM_EX_OK);
}
mainloop = g_main_loop_new(NULL, FALSE);
mainloop_add_signal(SIGTERM, mon_shutdown);
mainloop_add_signal(SIGINT, mon_shutdown);
#if CURSES_ENABLED
if (output_format == mon_output_console) {
ncurses_winch_handler = crm_signal_handler(SIGWINCH, mon_winresize);
if (ncurses_winch_handler == SIG_DFL ||
ncurses_winch_handler == SIG_IGN || ncurses_winch_handler == SIG_ERR)
ncurses_winch_handler = NULL;
io_channel = g_io_channel_unix_new(STDIN_FILENO);
g_io_add_watch(io_channel, G_IO_IN, detect_user_input, NULL);
}
#endif
refresh_trigger = mainloop_add_trigger(G_PRIORITY_LOW, mon_refresh_display, NULL);
g_main_loop_run(mainloop);
g_main_loop_unref(mainloop);
if (io_channel != NULL) {
g_io_channel_shutdown(io_channel, TRUE, NULL);
}
crm_info("Exiting %s", crm_system_name);
return clean_up(CRM_EX_OK);
}
/*!
* \internal
* \brief Print one-line status suitable for use with monitoring software
*
* \param[in] data_set Working set of CIB state
* \param[in] history List of stonith actions
*
* \note This function's output (and the return code when the program exits)
* should conform to https://www.monitoring-plugins.org/doc/guidelines.html
*/
static void
print_simple_status(pcmk__output_t *out, pe_working_set_t * data_set,
stonith_history_t *history, unsigned int mon_ops)
{
GListPtr gIter = NULL;
int nodes_online = 0;
int nodes_standby = 0;
int nodes_maintenance = 0;
char *offline_nodes = NULL;
gboolean no_dc = FALSE;
gboolean offline = FALSE;
if (data_set->dc_node == NULL) {
mon_ops |= mon_op_has_warnings;
no_dc = TRUE;
}
for (gIter = data_set->nodes; gIter != NULL; gIter = gIter->next) {
pe_node_t *node = (pe_node_t *) gIter->data;
if (node->details->standby && node->details->online) {
nodes_standby++;
} else if (node->details->maintenance && node->details->online) {
nodes_maintenance++;
} else if (node->details->online) {
nodes_online++;
} else {
char *s = crm_strdup_printf("offline node: %s", node->details->uname);
/* coverity[leaked_storage] False positive */
offline_nodes = pcmk__add_word(offline_nodes, s);
free(s);
mon_ops |= mon_op_has_warnings;
offline = TRUE;
}
}
if (is_set(mon_ops, mon_op_has_warnings)) {
out->info(out, "CLUSTER WARN:%s%s%s",
no_dc ? " No DC" : "",
no_dc && offline ? "," : "",
offline ? offline_nodes : "");
free(offline_nodes);
} else {
char *nodes_standby_s = NULL;
char *nodes_maint_s = NULL;
if (nodes_standby > 0) {
nodes_standby_s = crm_strdup_printf(", %d standby node%s", nodes_standby,
pcmk__plural_s(nodes_standby));
}
if (nodes_maintenance > 0) {
nodes_maint_s = crm_strdup_printf(", %d maintenance node%s",
nodes_maintenance,
pcmk__plural_s(nodes_maintenance));
}
out->info(out, "CLUSTER OK: %d node%s online%s%s, "
"%d resource instance%s configured",
nodes_online, pcmk__plural_s(nodes_online),
nodes_standby_s != NULL ? nodes_standby_s : "",
nodes_maint_s != NULL ? nodes_maint_s : "",
data_set->ninstances, pcmk__plural_s(data_set->ninstances));
free(nodes_standby_s);
free(nodes_maint_s);
}
/* coverity[leaked_storage] False positive */
}
/*!
* \internal
* \brief Reduce the stonith-history
* for successful actions we keep the last of every action-type & target
* for failed actions we record as well who had failed
* for actions in progress we keep full track
*
* \param[in] history List of stonith actions
*
*/
static stonith_history_t *
reduce_stonith_history(stonith_history_t *history)
{
stonith_history_t *new = history, *hp, *np;
if (new) {
hp = new->next;
new->next = NULL;
while (hp) {
stonith_history_t *hp_next = hp->next;
hp->next = NULL;
for (np = new; ; np = np->next) {
if ((hp->state == st_done) || (hp->state == st_failed)) {
/* action not in progress */
if (safe_str_eq(hp->target, np->target) &&
safe_str_eq(hp->action, np->action) &&
(hp->state == np->state) &&
((hp->state == st_done) ||
safe_str_eq(hp->delegate, np->delegate))) {
/* purge older hp */
stonith_history_free(hp);
break;
}
}
if (!np->next) {
np->next = hp;
break;
}
}
hp = hp_next;
}
}
return new;
}
static int
send_custom_trap(const char *node, const char *rsc, const char *task, int target_rc, int rc,
int status, const char *desc)
{
pid_t pid;
/*setenv needs chars, these are ints */
char *rc_s = crm_itoa(rc);
char *status_s = crm_itoa(status);
char *target_rc_s = crm_itoa(target_rc);
crm_debug("Sending external notification to '%s' via '%s'", options.external_recipient, options.external_agent);
if(rsc) {
setenv("CRM_notify_rsc", rsc, 1);
}
if (options.external_recipient) {
setenv("CRM_notify_recipient", options.external_recipient, 1);
}
setenv("CRM_notify_node", node, 1);
setenv("CRM_notify_task", task, 1);
setenv("CRM_notify_desc", desc, 1);
setenv("CRM_notify_rc", rc_s, 1);
setenv("CRM_notify_target_rc", target_rc_s, 1);
setenv("CRM_notify_status", status_s, 1);
pid = fork();
if (pid == -1) {
crm_perror(LOG_ERR, "notification fork() failed.");
}
if (pid == 0) {
/* crm_debug("notification: I am the child. Executing the nofitication program."); */
execl(options.external_agent, options.external_agent, NULL);
exit(CRM_EX_ERROR);
}
crm_trace("Finished running custom notification program '%s'.", options.external_agent);
free(target_rc_s);
free(status_s);
free(rc_s);
return 0;
}
static void
handle_rsc_op(xmlNode * xml, const char *node_id)
{
int rc = -1;
int status = -1;
int target_rc = -1;
gboolean notify = TRUE;
char *rsc = NULL;
char *task = NULL;
const char *desc = NULL;
const char *magic = NULL;
const char *id = NULL;
const char *node = NULL;
xmlNode *n = xml;
xmlNode * rsc_op = xml;
if(strcmp((const char*)xml->name, XML_LRM_TAG_RSC_OP) != 0) {
xmlNode *cIter;
for(cIter = xml->children; cIter; cIter = cIter->next) {
handle_rsc_op(cIter, node_id);
}
return;
}
id = crm_element_value(rsc_op, XML_LRM_ATTR_TASK_KEY);
if (id == NULL) {
/* Compatibility with <= 1.1.5 */
id = ID(rsc_op);
}
magic = crm_element_value(rsc_op, XML_ATTR_TRANSITION_MAGIC);
if (magic == NULL) {
/* non-change */
return;
}
if (!decode_transition_magic(magic, NULL, NULL, NULL, &status, &rc,
&target_rc)) {
crm_err("Invalid event %s detected for %s", magic, id);
return;
}
if (parse_op_key(id, &rsc, &task, NULL) == FALSE) {
crm_err("Invalid event detected for %s", id);
goto bail;
}
node = crm_element_value(rsc_op, XML_LRM_ATTR_TARGET);
while (n != NULL && safe_str_neq(XML_CIB_TAG_STATE, TYPE(n))) {
n = n->parent;
}
if(node == NULL && n) {
node = crm_element_value(n, XML_ATTR_UNAME);
}
if (node == NULL && n) {
node = ID(n);
}
if (node == NULL) {
node = node_id;
}
if (node == NULL) {
crm_err("No node detected for event %s (%s)", magic, id);
goto bail;
}
/* look up where we expected it to be? */
desc = pcmk_strerror(pcmk_ok);
if (status == PCMK_LRM_OP_DONE && target_rc == rc) {
crm_notice("%s of %s on %s completed: %s", task, rsc, node, desc);
if (rc == PCMK_OCF_NOT_RUNNING) {
notify = FALSE;
}
} else if (status == PCMK_LRM_OP_DONE) {
desc = services_ocf_exitcode_str(rc);
crm_warn("%s of %s on %s failed: %s", task, rsc, node, desc);
} else {
desc = services_lrm_status_str(status);
crm_warn("%s of %s on %s failed: %s", task, rsc, node, desc);
}
if (notify && options.external_agent) {
send_custom_trap(node, rsc, task, target_rc, rc, status, desc);
}
bail:
free(rsc);
free(task);
}
static gboolean
mon_trigger_refresh(gpointer user_data)
{
mainloop_set_trigger(refresh_trigger);
return FALSE;
}
#define NODE_PATT "/lrm[@id="
static char *
get_node_from_xpath(const char *xpath)
{
char *nodeid = NULL;
char *tmp = strstr(xpath, NODE_PATT);
if(tmp) {
tmp += strlen(NODE_PATT);
tmp += 1;
nodeid = strdup(tmp);
tmp = strstr(nodeid, "\'");
CRM_ASSERT(tmp);
tmp[0] = 0;
}
return nodeid;
}
static void
crm_diff_update_v2(const char *event, xmlNode * msg)
{
xmlNode *change = NULL;
xmlNode *diff = get_message_xml(msg, F_CIB_UPDATE_RESULT);
for (change = __xml_first_child(diff); change != NULL; change = __xml_next(change)) {
const char *name = NULL;
const char *op = crm_element_value(change, XML_DIFF_OP);
const char *xpath = crm_element_value(change, XML_DIFF_PATH);
xmlNode *match = NULL;
const char *node = NULL;
if(op == NULL) {
continue;
} else if(strcmp(op, "create") == 0) {
match = change->children;
} else if(strcmp(op, "move") == 0) {
continue;
} else if(strcmp(op, "delete") == 0) {
continue;
} else if(strcmp(op, "modify") == 0) {
match = first_named_child(change, XML_DIFF_RESULT);
if(match) {
match = match->children;
}
}
if(match) {
name = (const char *)match->name;
}
crm_trace("Handling %s operation for %s %p, %s", op, xpath, match, name);
if(xpath == NULL) {
/* Version field, ignore */
} else if(name == NULL) {
crm_debug("No result for %s operation to %s", op, xpath);
CRM_ASSERT(strcmp(op, "delete") == 0 || strcmp(op, "move") == 0);
} else if(strcmp(name, XML_TAG_CIB) == 0) {
xmlNode *state = NULL;
xmlNode *status = first_named_child(match, XML_CIB_TAG_STATUS);
for (state = __xml_first_child_element(status); state != NULL;
state = __xml_next_element(state)) {
node = crm_element_value(state, XML_ATTR_UNAME);
if (node == NULL) {
node = ID(state);
}
handle_rsc_op(state, node);
}
} else if(strcmp(name, XML_CIB_TAG_STATUS) == 0) {
xmlNode *state = NULL;
for (state = __xml_first_child_element(match); state != NULL;
state = __xml_next_element(state)) {
node = crm_element_value(state, XML_ATTR_UNAME);
if (node == NULL) {
node = ID(state);
}
handle_rsc_op(state, node);
}
} else if(strcmp(name, XML_CIB_TAG_STATE) == 0) {
node = crm_element_value(match, XML_ATTR_UNAME);
if (node == NULL) {
node = ID(match);
}
handle_rsc_op(match, node);
} else if(strcmp(name, XML_CIB_TAG_LRM) == 0) {
node = ID(match);
handle_rsc_op(match, node);
} else if(strcmp(name, XML_LRM_TAG_RESOURCES) == 0) {
char *local_node = get_node_from_xpath(xpath);
handle_rsc_op(match, local_node);
free(local_node);
} else if(strcmp(name, XML_LRM_TAG_RESOURCE) == 0) {
char *local_node = get_node_from_xpath(xpath);
handle_rsc_op(match, local_node);
free(local_node);
} else if(strcmp(name, XML_LRM_TAG_RSC_OP) == 0) {
char *local_node = get_node_from_xpath(xpath);
handle_rsc_op(match, local_node);
free(local_node);
} else {
crm_trace("Ignoring %s operation for %s %p, %s", op, xpath, match, name);
}
}
}
static void
crm_diff_update_v1(const char *event, xmlNode * msg)
{
/* Process operation updates */
xmlXPathObject *xpathObj = xpath_search(msg,
"//" F_CIB_UPDATE_RESULT "//" XML_TAG_DIFF_ADDED
"//" XML_LRM_TAG_RSC_OP);
int lpc = 0, max = numXpathResults(xpathObj);
for (lpc = 0; lpc < max; lpc++) {
xmlNode *rsc_op = getXpathResult(xpathObj, lpc);
handle_rsc_op(rsc_op, NULL);
}
freeXpathObject(xpathObj);
}
static void
crm_diff_update(const char *event, xmlNode * msg)
{
int rc = -1;
static bool stale = FALSE;
gboolean cib_updated = FALSE;
xmlNode *diff = get_message_xml(msg, F_CIB_UPDATE_RESULT);
print_dot(output_format);
if (current_cib != NULL) {
rc = xml_apply_patchset(current_cib, diff, TRUE);
switch (rc) {
case -pcmk_err_diff_resync:
case -pcmk_err_diff_failed:
crm_notice("[%s] Patch aborted: %s (%d)", event, pcmk_strerror(rc), rc);
free_xml(current_cib); current_cib = NULL;
break;
case pcmk_ok:
cib_updated = TRUE;
break;
default:
crm_notice("[%s] ABORTED: %s (%d)", event, pcmk_strerror(rc), rc);
free_xml(current_cib); current_cib = NULL;
}
}
if (current_cib == NULL) {
crm_trace("Re-requesting the full cib");
cib->cmds->query(cib, NULL, &current_cib, cib_scope_local | cib_sync_call);
}
if (options.external_agent) {
int format = 0;
crm_element_value_int(diff, "format", &format);
switch(format) {
case 1:
crm_diff_update_v1(event, msg);
break;
case 2:
crm_diff_update_v2(event, msg);
break;
default:
crm_err("Unknown patch format: %d", format);
}
}
if (current_cib == NULL) {
if(!stale) {
print_as(output_format, "--- Stale data ---");
}
stale = TRUE;
return;
}
stale = FALSE;
kick_refresh(cib_updated);
}
static gboolean
mon_refresh_display(gpointer user_data)
{
xmlNode *cib_copy = copy_xml(current_cib);
stonith_history_t *stonith_history = NULL;
int history_rc = 0;
last_refresh = time(NULL);
if (cli_config_update(&cib_copy, NULL, FALSE) == FALSE) {
if (cib) {
cib->cmds->signoff(cib);
}
out->err(out, "Upgrade failed: %s", pcmk_strerror(-pcmk_err_schema_validation));
clean_up(CRM_EX_CONFIG);
return FALSE;
}
/* get the stonith-history if there is evidence we need it
*/
while (is_set(options.mon_ops, mon_op_fence_history)) {
if (st != NULL) {
history_rc = st->cmds->history(st, st_opt_sync_call, NULL, &stonith_history, 120);
if (history_rc != 0) {
out->err(out, "Critical: Unable to get stonith-history");
mon_cib_connection_destroy_error(NULL);
} else {
stonith_history = stonith__sort_history(stonith_history);
if (is_not_set(options.mon_ops, mon_op_fence_full_history) && output_format != mon_output_xml) {
stonith_history = reduce_stonith_history(stonith_history);
}
break; /* all other cases are errors */
}
} else {
out->err(out, "Critical: No stonith-API");
}
free_xml(cib_copy);
out->err(out, "Reading stonith-history failed");
return FALSE;
}
if (mon_data_set == NULL) {
mon_data_set = pe_new_working_set();
CRM_ASSERT(mon_data_set != NULL);
}
set_bit(mon_data_set->flags, pe_flag_no_compat);
mon_data_set->input = cib_copy;
cluster_status(mon_data_set);
/* Unpack constraints if any section will need them
* (tickets may be referenced in constraints but not granted yet,
* and bans need negative location constraints) */
if (is_set(show, mon_show_bans) || is_set(show, mon_show_tickets)) {
xmlNode *cib_constraints = get_object_root(XML_CIB_TAG_CONSTRAINTS,
mon_data_set->input);
unpack_constraints(cib_constraints, mon_data_set);
}
switch (output_format) {
case mon_output_html:
case mon_output_cgi:
if (print_html_status(out, mon_data_set, stonith_history, options.mon_ops,
show, options.neg_location_prefix,
options.only_node) != 0) {
g_set_error(&error, G_OPTION_ERROR, CRM_EX_CANTCREAT, "Critical: Unable to output html file");
clean_up(CRM_EX_CANTCREAT);
return FALSE;
}
break;
case mon_output_legacy_xml:
case mon_output_xml:
print_xml_status(out, mon_data_set, crm_errno2exit(history_rc),
stonith_history, options.mon_ops, show,
options.neg_location_prefix, options.only_node);
break;
case mon_output_monitor:
print_simple_status(out, mon_data_set, stonith_history, options.mon_ops);
if (is_set(options.mon_ops, mon_op_has_warnings)) {
clean_up(MON_STATUS_WARN);
return FALSE;
}
break;
case mon_output_console:
/* If curses is not enabled, this will just fall through to the plain
* text case.
*/
#if CURSES_ENABLED
blank_screen();
print_status(out, mon_data_set, stonith_history, options.mon_ops, show,
options.neg_location_prefix, options.only_node);
refresh();
break;
#endif
case mon_output_plain:
print_status(out, mon_data_set, stonith_history, options.mon_ops, show,
options.neg_location_prefix, options.only_node);
break;
case mon_output_unset:
case mon_output_none:
break;
}
stonith_history_free(stonith_history);
stonith_history = NULL;
pe_reset_working_set(mon_data_set);
return TRUE;
}
static void
mon_st_callback_event(stonith_t * st, stonith_event_t * e)
{
if (st->state == stonith_disconnected) {
/* disconnect cib as well and have everything reconnect */
mon_cib_connection_destroy_regular(NULL);
} else if (options.external_agent) {
char *desc = crm_strdup_printf("Operation %s requested by %s for peer %s: %s (ref=%s)",
e->operation, e->origin, e->target, pcmk_strerror(e->result),
e->id);
send_custom_trap(e->target, NULL, e->operation, pcmk_ok, e->result, 0, desc);
free(desc);
}
}
static void
kick_refresh(gboolean data_updated)
{
static int updates = 0;
time_t now = time(NULL);
if (data_updated) {
updates++;
}
if(refresh_timer == NULL) {
refresh_timer = mainloop_timer_add("refresh", 2000, FALSE, mon_trigger_refresh, NULL);
}
/* Refresh
* - immediately if the last update was more than 5s ago
* - every 10 cib-updates
* - at most 2s after the last update
*/
if ((now - last_refresh) > (options.reconnect_msec / 1000)) {
mainloop_set_trigger(refresh_trigger);
mainloop_timer_stop(refresh_timer);
updates = 0;
} else if(updates >= 10) {
mainloop_set_trigger(refresh_trigger);
mainloop_timer_stop(refresh_timer);
updates = 0;
} else {
mainloop_timer_start(refresh_timer);
}
}
static void
mon_st_callback_display(stonith_t * st, stonith_event_t * e)
{
if (st->state == stonith_disconnected) {
/* disconnect cib as well and have everything reconnect */
mon_cib_connection_destroy_regular(NULL);
} else {
print_dot(output_format);
kick_refresh(TRUE);
}
}
static void
clean_up_connections(void)
{
if (cib != NULL) {
cib->cmds->signoff(cib);
cib_delete(cib);
cib = NULL;
}
if (st != NULL) {
if (st->state != stonith_disconnected) {
st->cmds->remove_notification(st, T_STONITH_NOTIFY_DISCONNECT);
st->cmds->remove_notification(st, T_STONITH_NOTIFY_FENCE);
st->cmds->remove_notification(st, T_STONITH_NOTIFY_HISTORY);
st->cmds->disconnect(st);
}
stonith_api_delete(st);
st = NULL;
}
}
static void
handle_html_output(crm_exit_t exit_code) {
xmlNodePtr html = NULL;
pcmk__html_add_header(html, "meta", "http-equiv", "refresh", "content",
crm_itoa(options.reconnect_msec/1000), NULL);
out->finish(out, exit_code, true, (void **) &html);
}
/*
* De-init ncurses, disconnect from the CIB manager, disconnect fencing,
* deallocate memory and show usage-message if requested.
*
* We don't actually return, but nominally returning crm_exit_t allows a usage
* like "return clean_up(exit_code);" which helps static analysis understand the
* code flow.
*/
static crm_exit_t
clean_up(crm_exit_t exit_code)
{
/* Quitting crm_mon is much more complicated than it ought to be. */
/* (1) Close connections, free things, etc. */
clean_up_connections();
free(options.pid_file);
free(options.neg_location_prefix);
g_slist_free_full(options.includes_excludes, free);
pe_free_working_set(mon_data_set);
mon_data_set = NULL;
g_strfreev(processed_args);
/* (2) If this is abnormal termination and we're in curses mode, shut down
* curses first. Any messages displayed to the screen before curses is shut
* down will be lost because doing the shut down will also restore the
* screen to whatever it looked like before crm_mon was started.
*/
if ((error != NULL || exit_code == CRM_EX_USAGE) && output_format == mon_output_console) {
out->finish(out, exit_code, false, NULL);
pcmk__output_free(out);
out = NULL;
}
/* (3) If this is a command line usage related failure, print the usage
* message.
*/
if (exit_code == CRM_EX_USAGE && (output_format == mon_output_console || output_format == mon_output_plain)) {
char *help = g_option_context_get_help(context, TRUE, NULL);
fprintf(stderr, "%s", help);
g_free(help);
}
pcmk__free_arg_context(context);
/* (4) If this is any kind of error, print the error out and exit. Make
* sure to handle situations both before and after formatted output is
* set up. We want errors to appear formatted if at all possible.
*/
if (error != NULL) {
if (out != NULL) {
out->err(out, "%s: %s", g_get_prgname(), error->message);
out->finish(out, exit_code, true, NULL);
pcmk__output_free(out);
} else {
fprintf(stderr, "%s: %s\n", g_get_prgname(), error->message);
}
g_clear_error(&error);
crm_exit(exit_code);
}
/* (5) Print formatted output to the screen if we made it far enough in
* crm_mon to be able to do so.
*/
if (out != NULL) {
switch (output_format) {
case mon_output_cgi:
case mon_output_html:
handle_html_output(exit_code);
break;
default:
out->finish(out, exit_code, true, NULL);
break;
}
pcmk__output_free(out);
pcmk__unregister_formats();
}
crm_exit(exit_code);
}

File Metadata

Mime Type
text/x-diff
Expires
Sat, Nov 23, 7:00 AM (1 d, 14 h)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
1018376
Default Alt Text
(274 KB)

Event Timeline