gboolean
crmd_authorize_message(xmlNode * client_msg, crmd_client_t * curr_client)
{
/* check the best case first */
const char *sys_from = crm_element_value(client_msg, F_CRM_SYS_FROM);
char *uuid = NULL;
char *client_name = NULL;
char *major_version = NULL;
char *minor_version = NULL;
const char *filtered_from;
gpointer table_key = NULL;
gboolean auth_result = FALSE;
struct crm_subsystem_s *the_subsystem = NULL;
gboolean can_reply = FALSE; /* no-one has registered with this id */
xmlNode *xml = NULL;
const char *op = crm_element_value(client_msg, F_CRM_TASK);
if (safe_str_neq(CRM_OP_HELLO, op)) {
if (sys_from == NULL) {
crm_warn("Message [%s] was had no value for %s... discarding",
crm_element_value(client_msg, XML_ATTR_REFERENCE), F_CRM_SYS_FROM);
return FALSE;
}
filtered_from = sys_from;
/* The CIB can have two names on the DC */
if (strcasecmp(sys_from, CRM_SYSTEM_DCIB) == 0)
filtered_from = CRM_SYSTEM_CIB;
if (g_hash_table_lookup(ipc_clients, filtered_from) != NULL) {
can_reply = TRUE; /* reply can be routed */
}
crm_trace("Message reply can%s be routed from %s.", can_reply ? "" : " not", sys_from);
if (can_reply == FALSE) {
crm_warn("Message [%s] not authorized",
crm_element_value(client_msg, XML_ATTR_REFERENCE));
}
return can_reply;
}
crm_trace("received client join msg");
crm_log_xml_trace(client_msg, "join");
xml = get_message_xml(client_msg, F_CRM_DATA);
auth_result = process_hello_message(xml, &uuid, &client_name, &major_version, &minor_version);
if (auth_result == TRUE) {
if (client_name == NULL || uuid == NULL) {
crm_err("Bad client details (client_name=%s, uuid=%s)",
crm_str(client_name), crm_str(uuid));
auth_result = FALSE;
}
}
if (auth_result == TRUE) {
/* check version */
int mav = atoi(major_version);
int miv = atoi(minor_version);
crm_trace("Checking client version number");
if (mav < 0 || miv < 0) {
crm_err("Client version (%d:%d) is not acceptable", mav, miv);
auth_result = FALSE;
}
crm_free(major_version);
crm_free(minor_version);
}
if (safe_str_eq(CRM_SYSTEM_PENGINE, client_name)) {
the_subsystem = pe_subsystem;
} else if (safe_str_eq(CRM_SYSTEM_TENGINE, client_name)) {
the_subsystem = te_subsystem;
}
/* TODO: Is this code required anymore?? */
if (auth_result == TRUE && the_subsystem != NULL) {
/* if we already have one of those clients
* only applies to te, pe etc. not admin clients
*/
crm_err("Checking if %s is required/already connected", client_name);
table_key = (gpointer) crm_strdup(client_name);
if (is_set(fsa_input_register, the_subsystem->flag_connected)) {
auth_result = FALSE;
crm_free(table_key);
table_key = NULL;
crm_warn("Bit\t%.16llx set in %.16llx",
the_subsystem->flag_connected, fsa_input_register);
crm_err("Client %s is already connected", client_name);
} else if (FALSE == is_set(fsa_input_register, the_subsystem->flag_required)) {
crm_warn("Bit\t%.16llx not set in %.16llx",
the_subsystem->flag_connected, fsa_input_register);
crm_warn("Client %s joined but we dont need it", client_name);
stop_subsystem(the_subsystem, TRUE);
} else {
the_subsystem->ipc = curr_client->client_channel;
set_bit_inplace(fsa_input_register, the_subsystem->flag_connected);
}
} else {
table_key = (gpointer) generate_hash_key(client_name, uuid);
}
if (auth_result == TRUE) {
crm_trace("Accepted client %s", crm_str(table_key));
curr_client->table_key = table_key;
curr_client->sub_sys = crm_strdup(client_name);
curr_client->uuid = crm_strdup(uuid);
g_hash_table_insert(ipc_clients, table_key, curr_client->client_channel);
send_hello_message(curr_client->client_channel, "n/a", CRM_SYSTEM_CRMD, "0", "1");
crm_trace("Updated client list with %s", crm_str(table_key));
crm_trace("Triggering FSA: %s", __FUNCTION__);
mainloop_set_trigger(fsa_source);
if (the_subsystem != NULL) {
CRM_CHECK(the_subsystem->client == NULL,
process_client_disconnect(the_subsystem->client));
the_subsystem->client = curr_client;
}
} else {
crm_free(table_key);
crm_warn("Rejected client logon request");
curr_client->client_channel->ch_status = IPC_DISC_PENDING;
}
if (uuid != NULL)
crm_free(uuid);
if (minor_version != NULL)
crm_free(minor_version);
if (major_version != NULL)
crm_free(major_version);
if (client_name != NULL)
crm_free(client_name);
/* hello messages should never be processed further */
return FALSE;
}
/* A_ELECTION_COUNT */
void
do_election_count_vote(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
int age = 0;
int election_id = -1;
int log_level = LOG_INFO;
gboolean use_born_on = FALSE;
gboolean done = FALSE;
gboolean we_loose = FALSE;
const char *op = NULL;
const char *vote_from = NULL;
const char *election_owner = NULL;
const char *reason = "unknown";
crm_node_t *our_node = NULL, *your_node = NULL;
ha_msg_input_t *vote = fsa_typed_data(fsa_dt_ha_msg);
static int election_wins = 0;
time_t tm_now = time(NULL);
static time_t expires = 0;
static time_t last_election_loss = 0;
/* if the membership copy is NULL we REALLY shouldnt be voting
* the question is how we managed to get here.
*/
CRM_CHECK(msg_data != NULL, return);
CRM_CHECK(vote != NULL, crm_err("Bogus data from %s", msg_data->origin); return);
CRM_CHECK(vote->msg != NULL, crm_err("Bogus data from %s", msg_data->origin); return);
if(crm_peer_cache == NULL) {
if(is_not_set(fsa_input_register, R_SHUTDOWN)) {
crm_err("Internal error, no peer cache");
}
return;
}
op = crm_element_value(vote->msg, F_CRM_TASK);
vote_from = crm_element_value(vote->msg, F_CRM_HOST_FROM);
election_owner = crm_element_value(vote->msg, F_CRM_ELECTION_OWNER);
crm_element_value_int(vote->msg, F_CRM_ELECTION_ID, &election_id);
CRM_CHECK(vote_from != NULL, vote_from = fsa_our_uname);
your_node = crm_get_peer(0, vote_from);
our_node = crm_get_peer(0, fsa_our_uname);
if (voted == NULL) {
crm_debug("Created voted hash");
voted = g_hash_table_new_full(crm_str_hash, g_str_equal,
g_hash_destroy_str, g_hash_destroy_str);
}
if (is_heartbeat_cluster()) {
use_born_on = TRUE;
} else if (is_classic_ais_cluster()) {
use_born_on = TRUE;
}
if (cur_state == S_STARTING) {
reason = "Still starting";
we_loose = TRUE;
} else if (our_node == NULL || crm_is_peer_active(our_node) == FALSE) {
reason = "We are not part of the cluster";
log_level = LOG_ERR;
we_loose = TRUE;
} else if (election_id != current_election_id && crm_str_eq(fsa_our_uuid, election_owner, TRUE)) {
log_level = LOG_DEBUG_2;
reason = "Superceeded";
done = TRUE;
} else if (your_node == NULL || crm_is_peer_active(your_node) == FALSE) {
/* Possibly we cached the message in the FSA queue at a point that it wasn't */
reason = "Peer is not part of our cluster";
log_level = LOG_WARNING;
done = TRUE;
} else if (crm_str_eq(op, CRM_OP_NOVOTE, TRUE)) {
char *op_copy = strdup(op);
char *uname_copy = strdup(vote_from);
CRM_ASSERT(crm_str_eq(fsa_our_uuid, election_owner, TRUE));
/* update the list of nodes that have voted */
g_hash_table_replace(voted, uname_copy, op_copy);
reason = "Recorded";
done = TRUE;
} else {
struct timeval your_age;
const char *your_version = crm_element_value(vote->msg, F_CRM_VERSION);
your_age.tv_sec = 0;
your_age.tv_usec = 0;
crm_element_value_int(vote->msg, F_CRM_ELECTION_AGE_S, (int *)&(your_age.tv_sec));
crm_element_value_int(vote->msg, F_CRM_ELECTION_AGE_US, (int *)&(your_age.tv_usec));
age = crm_compare_age(your_age);
if(your_age.tv_sec == 0 && your_age.tv_usec == 0) {
crm_log_xml_trace(vote->msg, "bad vote");
crm_write_blackbox(0, NULL);
}
if (crm_str_eq(vote_from, fsa_our_uname, TRUE)) {
char *op_copy = strdup(op);
char *uname_copy = strdup(vote_from);
CRM_ASSERT(crm_str_eq(fsa_our_uuid, election_owner, TRUE));
/* update ourselves in the list of nodes that have voted */
g_hash_table_replace(voted, uname_copy, op_copy);
reason = "Recorded";
done = TRUE;
} else if (compare_version(your_version, CRM_FEATURE_SET) < 0) {
reason = "Version";
we_loose = TRUE;
} else if (compare_version(your_version, CRM_FEATURE_SET) > 0) {
reason = "Version";
} else if (age < 0) {
reason = "Uptime";
we_loose = TRUE;
} else if (age > 0) {
reason = "Uptime";
/* TODO: Check for y(our) born < 0 */
} else if (use_born_on && your_node->born < our_node->born) {
reason = "Born";
we_loose = TRUE;
} else if (use_born_on && your_node->born > our_node->born) {
reason = "Born";
} else if (fsa_our_uname == NULL) {
reason = "Unknown host name";
we_loose = TRUE;
} else if (strcasecmp(fsa_our_uname, vote_from) > 0) {
reason = "Host name";
we_loose = TRUE;
} else {
reason = "Host name";
CRM_ASSERT(strcasecmp(fsa_our_uname, vote_from) < 0);
/* cant happen...
* } else if(strcasecmp(fsa_our_uname, vote_from) == 0) {
*
*/
}
}
if (expires < tm_now) {
election_wins = 0;
expires = tm_now + STORM_INTERVAL;
} else if (done == FALSE && we_loose == FALSE) {
int peers = 1 + g_hash_table_size(crm_peer_cache);
/* If every node has to vote down every other node, thats N*(N-1) total elections
* Allow some leway before _really_ complaining
*/
election_wins++;
if (election_wins > (peers * peers)) {
crm_warn("Election storm detected: %d elections in %d seconds", election_wins,
STORM_INTERVAL);
election_wins = 0;
expires = tm_now + STORM_INTERVAL;
crm_write_blackbox(0, NULL);
}
}
if (done) {
do_crm_log(log_level + 1, "Election %d (current: %d, owner: %s): Processed %s from %s (%s)",
election_id, current_election_id, election_owner, op, vote_from, reason);
} else if (we_loose) {
xmlNode *novote = create_request(CRM_OP_NOVOTE, NULL, vote_from,
CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL);
do_crm_log(log_level, "Election %d (owner: %s) lost: %s from %s (%s)",
election_id, election_owner, op, vote_from, reason);
update_dc(NULL);
crm_timer_stop(election_timeout);
if (fsa_input_register & R_THE_DC) {
crm_trace("Give up the DC to %s", vote_from);
register_fsa_input(C_FSA_INTERNAL, I_RELEASE_DC, NULL);
fsa_cib_conn->cmds->set_slave(fsa_cib_conn, cib_scope_local);
} else if (cur_state != S_STARTING) {
crm_trace("We werent the DC anyway");
register_fsa_input(C_FSA_INTERNAL, I_PENDING, NULL);
}
crm_xml_add(novote, F_CRM_ELECTION_OWNER, election_owner);
crm_xml_add_int(novote, F_CRM_ELECTION_ID, election_id);
send_cluster_message(crm_get_peer(0, vote_from), crm_msg_crmd, novote, TRUE);
free_xml(novote);
last_election_loss = tm_now;
} else {
do_crm_log(log_level, "Election %d (owner: %s) pass: %s from %s (%s)",
election_id, election_owner, op, vote_from, reason);
if (last_election_loss) {
if (tm_now - last_election_loss < (time_t) loss_dampen) {
crm_info("Election %d ignore: We already lost an election less than %ds ago (%s)",
election_id, loss_dampen, ctime(&last_election_loss));
update_dc(NULL);
return;
}
last_election_loss = 0;
}
register_fsa_input(C_FSA_INTERNAL, I_ELECTION, NULL);
g_hash_table_destroy(voted);
voted = NULL;
}
}
static char *
convert_non_atomic_uuid(char *old_uuid, resource_t * rsc, gboolean allow_notify,
gboolean free_original)
{
int interval = 0;
char *uuid = NULL;
char *rid = NULL;
char *raw_task = NULL;
int task = no_action;
crm_trace("Processing %s", old_uuid);
if (old_uuid == NULL) {
return NULL;
} else if (strstr(old_uuid, "notify") != NULL) {
goto done; /* no conversion */
} else if (rsc->variant < pe_group) {
goto done; /* no conversion */
}
CRM_ASSERT(parse_op_key(old_uuid, &rid, &raw_task, &interval));
if (interval > 0) {
goto done; /* no conversion */
}
task = text2task(raw_task);
switch (task) {
case stop_rsc:
case start_rsc:
case action_notify:
case action_promote:
case action_demote:
break;
case stopped_rsc:
case started_rsc:
case action_notified:
case action_promoted:
case action_demoted:
task--;
break;
case monitor_rsc:
case shutdown_crm:
case stonith_node:
task = no_action;
break;
default:
crm_err("Unknown action: %s", raw_task);
task = no_action;
break;
}
if (task != no_action) {
if (is_set(rsc->flags, pe_rsc_notify) && allow_notify) {
uuid = generate_notify_key(rid, "confirmed-post", task2text(task + 1));
} else {
uuid = generate_op_key(rid, task2text(task + 1), 0);
}
crm_trace("Converted %s -> %s", old_uuid, uuid);
}
done:
if (uuid == NULL) {
uuid = crm_strdup(old_uuid);
}
if (free_original) {
crm_free(old_uuid);
}
crm_free(raw_task);
crm_free(rid);
return uuid;
}
svc_action_t *
resources_action_create(const char *name, const char *standard, const char *provider,
const char *agent, const char *action, int interval, int timeout,
GHashTable * params, enum svc_action_flags flags)
{
svc_action_t *op = NULL;
/*
* Do some up front sanity checks before we go off and
* build the svc_action_t instance.
*/
if (crm_strlen_zero(name)) {
crm_err("A service or resource action must have a name.");
goto return_error;
}
if (crm_strlen_zero(standard)) {
crm_err("A service action must have a valid standard.");
goto return_error;
}
if (!strcasecmp(standard, "ocf") && crm_strlen_zero(provider)) {
crm_err("An OCF resource action must have a provider.");
goto return_error;
}
if (crm_strlen_zero(agent)) {
crm_err("A service or resource action must have an agent.");
goto return_error;
}
if (crm_strlen_zero(action)) {
crm_err("A service or resource action must specify an action.");
goto return_error;
}
if (safe_str_eq(action, "monitor") && (
#if SUPPORT_HEARTBEAT
safe_str_eq(standard, "heartbeat") ||
#endif
safe_str_eq(standard, "lsb") || safe_str_eq(standard, "service"))) {
action = "status";
}
/*
* Sanity checks passed, proceed!
*/
op = calloc(1, sizeof(svc_action_t));
op->opaque = calloc(1, sizeof(svc_action_private_t));
op->opaque->pending = NULL;
op->rsc = strdup(name);
op->action = strdup(action);
op->interval = interval;
op->timeout = timeout;
op->standard = strdup(standard);
op->agent = strdup(agent);
op->sequence = ++operations;
op->flags = flags;
if (asprintf(&op->id, "%s_%s_%d", name, action, interval) == -1) {
goto return_error;
}
if (strcasecmp(op->standard, "service") == 0) {
const char *expanded = resources_find_service_class(op->agent);
if(expanded) {
crm_debug("Found a %s agent for %s/%s", expanded, op->rsc, op->agent);
free(op->standard);
op->standard = strdup(expanded);
} else {
crm_info("Cannot determine the standard for %s (%s)", op->rsc, op->agent);
free(op->standard);
op->standard = strdup("lsb");
}
CRM_ASSERT(op->standard);
}
if (strcasecmp(op->standard, "ocf") == 0) {
op->provider = strdup(provider);
op->params = params;
params = NULL;
if (asprintf(&op->opaque->exec, "%s/resource.d/%s/%s", OCF_ROOT_DIR, provider, agent) == -1) {
crm_err("Internal error: cannot create agent path");
goto return_error;
}
op->opaque->args[0] = strdup(op->opaque->exec);
op->opaque->args[1] = strdup(action);
} else if (strcasecmp(op->standard, "lsb") == 0) {
if (op->agent[0] == '/') {
/* if given an absolute path, use that instead
* of tacking on the LSB_ROOT_DIR path to the front */
op->opaque->exec = strdup(op->agent);
} else if (asprintf(&op->opaque->exec, "%s/%s", LSB_ROOT_DIR, op->agent) == -1) {
crm_err("Internal error: cannot create agent path");
goto return_error;
}
op->opaque->args[0] = strdup(op->opaque->exec);
op->opaque->args[1] = strdup(op->action);
op->opaque->args[2] = NULL;
#if SUPPORT_HEARTBEAT
} else if (strcasecmp(op->standard, "heartbeat") == 0) {
int index;
int param_num;
char buf_tmp[20];
void *value_tmp;
if (op->agent[0] == '/') {
/* if given an absolute path, use that instead
* of tacking on the HB_RA_DIR path to the front */
op->opaque->exec = strdup(op->agent);
} else if (asprintf(&op->opaque->exec, "%s/%s", HB_RA_DIR, op->agent) == -1) {
crm_err("Internal error: cannot create agent path");
goto return_error;
}
op->opaque->args[0] = strdup(op->opaque->exec);
/* The "heartbeat" agent class only has positional arguments,
* which we keyed by their decimal position number. */
param_num = 1;
for (index = 1; index <= MAX_ARGC - 3; index++ ) {
snprintf(buf_tmp, sizeof(buf_tmp), "%d", index);
value_tmp = g_hash_table_lookup(params, buf_tmp);
if (value_tmp == NULL) {
/* maybe: strdup("") ??
* But the old lrmd did simply continue as well. */
continue;
}
op->opaque->args[param_num++] = strdup(value_tmp);
}
/* Add operation code as the last argument, */
/* and the teminating NULL pointer */
op->opaque->args[param_num++] = strdup(op->action);
op->opaque->args[param_num] = NULL;
#endif
#if SUPPORT_SYSTEMD
} else if (strcasecmp(op->standard, "systemd") == 0) {
op->opaque->exec = strdup("systemd-dbus");
#endif
#if SUPPORT_UPSTART
} else if (strcasecmp(op->standard, "upstart") == 0) {
op->opaque->exec = strdup("upstart-dbus");
#endif
} else if (strcasecmp(op->standard, "service") == 0) {
op->opaque->exec = strdup(SERVICE_SCRIPT);
op->opaque->args[0] = strdup(SERVICE_SCRIPT);
op->opaque->args[1] = strdup(agent);
op->opaque->args[2] = strdup(action);
#if SUPPORT_NAGIOS
} else if (strcasecmp(op->standard, "nagios") == 0) {
int index = 0;
if (op->agent[0] == '/') {
/* if given an absolute path, use that instead
* of tacking on the NAGIOS_PLUGIN_DIR path to the front */
op->opaque->exec = strdup(op->agent);
} else if (asprintf(&op->opaque->exec, "%s/%s", NAGIOS_PLUGIN_DIR, op->agent) == -1) {
crm_err("Internal error: cannot create agent path");
goto return_error;
}
op->opaque->args[0] = strdup(op->opaque->exec);
index = 1;
if (safe_str_eq(op->action, "monitor") && op->interval == 0) {
/* Invoke --version for a nagios probe */
op->opaque->args[index] = strdup("--version");
index++;
} else if (params) {
GHashTableIter iter;
char *key = NULL;
char *value = NULL;
static int args_size = sizeof(op->opaque->args) / sizeof(char *);
g_hash_table_iter_init(&iter, params);
while (g_hash_table_iter_next(&iter, (gpointer *) & key, (gpointer *) & value) &&
index <= args_size - 3) {
int len = 3;
char *long_opt = NULL;
if (safe_str_eq(key, XML_ATTR_CRM_VERSION) || strstr(key, CRM_META "_")) {
continue;
}
len += strlen(key);
long_opt = calloc(1, len);
sprintf(long_opt, "--%s", key);
long_opt[len - 1] = 0;
op->opaque->args[index] = long_opt;
op->opaque->args[index + 1] = strdup(value);
index += 2;
}
}
op->opaque->args[index] = NULL;
#endif
} else {
crm_err("Unknown resource standard: %s", op->standard);
services_action_free(op);
op = NULL;
}
if(params) {
g_hash_table_destroy(params);
}
return op;
return_error:
if(params) {
g_hash_table_destroy(params);
}
services_action_free(op);
return NULL;
}
/* A_TE_START, A_TE_STOP, A_TE_RESTART */
void
do_te_control(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
gboolean init_ok = TRUE;
if (action & A_TE_STOP) {
if (transition_graph) {
destroy_graph(transition_graph);
transition_graph = NULL;
}
if (fsa_cib_conn) {
fsa_cib_conn->cmds->del_notify_callback(
fsa_cib_conn, T_CIB_DIFF_NOTIFY, te_update_diff);
}
clear_bit(fsa_input_register, te_subsystem->flag_connected);
crm_info("Transitioner is now inactive");
}
if ((action & A_TE_START) == 0) {
return;
} else if (is_set(fsa_input_register, te_subsystem->flag_connected)) {
crm_debug("The transitioner is already active");
return;
} else if ((action & A_TE_START) && cur_state == S_STOPPING) {
crm_info("Ignoring request to start %s while shutting down", te_subsystem->name);
return;
}
te_uuid = crm_generate_uuid();
crm_info("Registering TE UUID: %s", te_uuid);
if (transition_trigger == NULL) {
transition_trigger = mainloop_add_trigger(G_PRIORITY_LOW, te_graph_trigger, NULL);
}
if (pcmk_ok !=
fsa_cib_conn->cmds->add_notify_callback(fsa_cib_conn, T_CIB_DIFF_NOTIFY, te_update_diff)) {
crm_err("Could not set CIB notification callback");
init_ok = FALSE;
}
if (pcmk_ok != fsa_cib_conn->cmds->set_op_callback(fsa_cib_conn, global_cib_callback)) {
crm_err("Could not set CIB global callback");
init_ok = FALSE;
}
if (init_ok) {
set_graph_functions(&te_graph_fns);
if (transition_graph) {
destroy_graph(transition_graph);
}
/* create a blank one */
crm_debug("Transitioner is now active");
transition_graph = create_blank_graph();
set_bit(fsa_input_register, te_subsystem->flag_connected);
}
}
int
main(int argc, char **argv, char **envp)
{
int flag = 0;
int index = 0;
int bump_log_num = 0;
const char *option = NULL;
/* If necessary, create PID1 now before any FDs are opened */
spawn_pidone(argc, argv, envp);
#ifndef ENABLE_PCMK_REMOTE
crm_log_preinit("pacemaker-execd", argc, argv);
crm_set_options(NULL, "[options]", long_options,
"Resource agent executor daemon for cluster nodes");
#else
crm_log_preinit("pacemaker-remoted", argc, argv);
crm_set_options(NULL, "[options]", long_options,
"Resource agent executor daemon for Pacemaker Remote nodes");
#endif
while (1) {
flag = crm_get_option(argc, argv, &index);
if (flag == -1) {
break;
}
switch (flag) {
case 'l':
crm_add_logfile(optarg);
break;
case 'p':
setenv("PCMK_remote_port", optarg, 1);
break;
case 'V':
bump_log_num++;
break;
case '?':
case '$':
crm_help(flag, CRM_EX_OK);
break;
default:
crm_help('?', CRM_EX_USAGE);
break;
}
}
crm_log_init(NULL, LOG_INFO, TRUE, FALSE, argc, argv, FALSE);
while (bump_log_num > 0) {
crm_bump_log_level(argc, argv);
bump_log_num--;
}
option = daemon_option("logfacility");
if(option && safe_str_neq(option, "none")) {
setenv("HA_LOGFACILITY", option, 1); /* Used by the ocf_log/ha_log OCF macro */
}
option = daemon_option("logfile");
if(option && safe_str_neq(option, "none")) {
setenv("HA_LOGFILE", option, 1); /* Used by the ocf_log/ha_log OCF macro */
if (daemon_option_enabled(crm_system_name, "debug")) {
setenv("HA_DEBUGLOG", option, 1); /* Used by the ocf_log/ha_debug OCF macro */
}
}
/* The presence of this variable allegedly controls whether child
* processes like httpd will try and use Systemd's sd_notify
* API
*/
unsetenv("NOTIFY_SOCKET");
/* Used by RAs - Leave owned by root */
crm_build_path(CRM_RSCTMP_DIR, 0755);
rsc_list = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free_rsc);
ipcs = mainloop_add_ipc_server(CRM_SYSTEM_LRMD, QB_IPC_SHM, &lrmd_ipc_callbacks);
if (ipcs == NULL) {
crm_err("Failed to create IPC server: shutting down and inhibiting respawn");
crm_exit(CRM_EX_FATAL);
}
#ifdef ENABLE_PCMK_REMOTE
if (lrmd_init_remote_tls_server() < 0) {
crm_err("Failed to create TLS listener: shutting down and staying down");
crm_exit(CRM_EX_FATAL);
}
ipc_proxy_init();
#endif
mainloop_add_signal(SIGTERM, lrmd_shutdown);
mainloop = g_main_loop_new(NULL, FALSE);
crm_info("Starting");
g_main_loop_run(mainloop);
/* should never get here */
lrmd_exit(NULL);
return CRM_EX_OK;
}
static void
action_synced_wait(svc_action_t * op, sigset_t mask)
{
#ifndef HAVE_SYS_SIGNALFD_H
CRM_ASSERT(FALSE);
#else
int status = 0;
int timeout = op->timeout;
int sfd = -1;
time_t start = -1;
struct pollfd fds[3];
int wait_rc = 0;
sfd = signalfd(-1, &mask, SFD_NONBLOCK);
if (sfd < 0) {
crm_perror(LOG_ERR, "signalfd() failed");
}
fds[0].fd = op->opaque->stdout_fd;
fds[0].events = POLLIN;
fds[0].revents = 0;
fds[1].fd = op->opaque->stderr_fd;
fds[1].events = POLLIN;
fds[1].revents = 0;
fds[2].fd = sfd;
fds[2].events = POLLIN;
fds[2].revents = 0;
crm_trace("Waiting for %d", op->pid);
start = time(NULL);
do {
int poll_rc = poll(fds, 3, timeout);
if (poll_rc > 0) {
if (fds[0].revents & POLLIN) {
svc_read_output(op->opaque->stdout_fd, op, FALSE);
}
if (fds[1].revents & POLLIN) {
svc_read_output(op->opaque->stderr_fd, op, TRUE);
}
if (fds[2].revents & POLLIN) {
struct signalfd_siginfo fdsi;
ssize_t s;
s = read(sfd, &fdsi, sizeof(struct signalfd_siginfo));
if (s != sizeof(struct signalfd_siginfo)) {
crm_perror(LOG_ERR, "Read from signal fd %d failed", sfd);
} else if (fdsi.ssi_signo == SIGCHLD) {
wait_rc = waitpid(op->pid, &status, WNOHANG);
if (wait_rc < 0){
crm_perror(LOG_ERR, "waitpid() for %d failed", op->pid);
} else if (wait_rc > 0) {
break;
}
}
}
} else if (poll_rc == 0) {
timeout = 0;
break;
} else if (poll_rc < 0) {
if (errno != EINTR) {
crm_perror(LOG_ERR, "poll() failed");
break;
}
}
timeout = op->timeout - (time(NULL) - start) * 1000;
} while ((op->timeout < 0 || timeout > 0));
crm_trace("Child done: %d", op->pid);
if (wait_rc <= 0) {
int killrc = kill(op->pid, SIGKILL);
op->rc = PCMK_OCF_UNKNOWN_ERROR;
if (op->timeout > 0 && timeout <= 0) {
op->status = PCMK_LRM_OP_TIMEOUT;
crm_warn("%s:%d - timed out after %dms", op->id, op->pid, op->timeout);
} else {
op->status = PCMK_LRM_OP_ERROR;
}
if (killrc && errno != ESRCH) {
crm_err("kill(%d, KILL) failed: %d", op->pid, errno);
}
/*
* From sigprocmask(2):
* It is not possible to block SIGKILL or SIGSTOP. Attempts to do so are silently ignored.
*
* This makes it safe to skip WNOHANG here
*/
waitpid(op->pid, &status, 0);
} else if (WIFEXITED(status)) {
op->status = PCMK_LRM_OP_DONE;
op->rc = WEXITSTATUS(status);
crm_info("Managed %s process %d exited with rc=%d", op->id, op->pid, op->rc);
} else if (WIFSIGNALED(status)) {
int signo = WTERMSIG(status);
op->status = PCMK_LRM_OP_ERROR;
crm_err("Managed %s process %d exited with signal=%d", op->id, op->pid, signo);
}
#ifdef WCOREDUMP
if (WCOREDUMP(status)) {
crm_err("Managed %s process %d dumped core", op->id, op->pid);
}
#endif
svc_read_output(op->opaque->stdout_fd, op, FALSE);
svc_read_output(op->opaque->stderr_fd, op, TRUE);
close(op->opaque->stdout_fd);
close(op->opaque->stderr_fd);
close(sfd);
#endif
}
int
main(int argc, char **argv)
{
int flag = 0;
int index = 0;
int argerr = 0;
crm_ipc_t *old_instance = NULL;
attrd_init_mainloop();
crm_log_preinit(NULL, argc, argv);
crm_set_options(NULL, "[options]", long_options,
"Daemon for aggregating and atomically storing node attribute updates into the CIB");
mainloop_add_signal(SIGTERM, attrd_shutdown);
while (1) {
flag = crm_get_option(argc, argv, &index);
if (flag == -1)
break;
switch (flag) {
case 'V':
crm_bump_log_level(argc, argv);
break;
case 'h': /* Help message */
crm_help(flag, CRM_EX_OK);
break;
default:
++argerr;
break;
}
}
if (optind > argc) {
++argerr;
}
if (argerr) {
crm_help('?', CRM_EX_USAGE);
}
crm_log_init(T_ATTRD, LOG_INFO, TRUE, FALSE, argc, argv, FALSE);
crm_notice("Starting Pacemaker node attribute manager");
old_instance = crm_ipc_new(T_ATTRD, 0);
if (crm_ipc_connect(old_instance)) {
/* IPC end-point already up */
crm_ipc_close(old_instance);
crm_ipc_destroy(old_instance);
crm_err("pacemaker-attrd is already active, aborting startup");
crm_exit(CRM_EX_OK);
} else {
/* not up or not authentic, we'll proceed either way */
crm_ipc_destroy(old_instance);
old_instance = NULL;
}
attributes = g_hash_table_new_full(crm_str_hash, g_str_equal, NULL, free_attribute);
/* Connect to the CIB before connecting to the cluster or listening for IPC.
* This allows us to assume the CIB is connected whenever we process a
* cluster or IPC message (which also avoids start-up race conditions).
*/
if (attrd_cib_connect(10) != pcmk_ok) {
attrd_exit_status = CRM_EX_FATAL;
goto done;
}
crm_info("CIB connection active");
if (attrd_cluster_connect() != pcmk_ok) {
attrd_exit_status = CRM_EX_FATAL;
goto done;
}
crm_info("Cluster connection active");
// Initialization that requires the cluster to be connected
attrd_election_init();
attrd_cib_init();
/* Set a private attribute for ourselves with the protocol version we
* support. This lets all nodes determine the minimum supported version
* across all nodes. It also ensures that the writer learns our node name,
* so it can send our attributes to the CIB.
*/
attrd_broadcast_protocol();
attrd_init_ipc(&ipcs, attrd_ipc_dispatch);
crm_notice("Pacemaker node attribute manager successfully started and accepting connections");
attrd_run_mainloop();
done:
crm_info("Shutting down attribute manager");
attrd_election_fini();
attrd_ipc_fini();
attrd_lrmd_disconnect();
attrd_cib_disconnect();
g_hash_table_destroy(attributes);
crm_exit(attrd_exit_status);
}
GList *
upstart_job_listall(void)
{
GList *units = NULL;
DBusMessageIter args;
DBusMessageIter unit;
DBusMessage *msg = NULL;
DBusMessage *reply = NULL;
const char *method = "GetAllJobs";
DBusError error;
int lpc = 0;
if (upstart_init() == FALSE) {
return NULL;
}
/*
com.ubuntu.Upstart0_6.GetAllJobs (out <Array of ObjectPath> jobs)
*/
dbus_error_init(&error);
msg = dbus_message_new_method_call(BUS_NAME, // target for the method call
BUS_PATH, // object to call on
UPSTART_06_API, // interface to call on
method); // method name
CRM_ASSERT(msg != NULL);
reply = pcmk_dbus_send_recv(msg, upstart_proxy, &error);
dbus_message_unref(msg);
if(error.name) {
crm_err("Call to %s failed: %s", method, error.name);
return NULL;
} else if (!dbus_message_iter_init(reply, &args)) {
crm_err("Call to %s failed: Message has no arguments", method);
dbus_message_unref(reply);
return NULL;
}
if(!pcmk_dbus_type_check(reply, &args, DBUS_TYPE_ARRAY, __FUNCTION__, __LINE__)) {
crm_err("Call to %s failed: Message has invalid arguments", method);
dbus_message_unref(reply);
return NULL;
}
dbus_message_iter_recurse(&args, &unit);
while (dbus_message_iter_get_arg_type (&unit) != DBUS_TYPE_INVALID) {
DBusBasicValue value;
const char *job = NULL;
char *path = NULL;
if(!pcmk_dbus_type_check(reply, &unit, DBUS_TYPE_OBJECT_PATH, __FUNCTION__, __LINE__)) {
continue;
}
dbus_message_iter_get_basic(&unit, &value);
if(value.str) {
int llpc = 0;
path = value.str;
job = value.str;
while (path[llpc] != 0) {
if (path[llpc] == '/') {
job = path + llpc + 1;
}
llpc++;
}
lpc++;
crm_trace("%s -> %s\n", path, job);
units = g_list_append(units, fix_upstart_name(job));
}
dbus_message_iter_next (&unit);
}
dbus_message_unref(reply);
crm_trace("Found %d upstart jobs", lpc);
return units;
}
static gboolean
mainloop_gio_callback(GIOChannel *gio, GIOCondition condition, gpointer data)
{
gboolean keep = TRUE;
mainloop_io_t *client = data;
if(condition & G_IO_IN) {
if(client->ipc) {
long rc = 0;
int max = 10;
do {
rc = crm_ipc_read(client->ipc);
if(rc <= 0) {
crm_trace("Message acquisition from %s[%p] failed: %s (%ld)",
client->name, client, pcmk_strerror(rc), rc);
} else if(client->dispatch_fn_ipc) {
const char *buffer = crm_ipc_buffer(client->ipc);
crm_trace("New message from %s[%p] = %d", client->name, client, rc, condition);
if(client->dispatch_fn_ipc(buffer, rc, client->userdata) < 0) {
crm_trace("Connection to %s no longer required", client->name);
keep = FALSE;
}
}
} while(keep && rc > 0 && --max > 0);
} else {
crm_trace("New message from %s[%p]", client->name, client);
if(client->dispatch_fn_io) {
if(client->dispatch_fn_io(client->userdata) < 0) {
crm_trace("Connection to %s no longer required", client->name);
keep = FALSE;
}
}
}
}
if(client->ipc && crm_ipc_connected(client->ipc) == FALSE) {
crm_err("Connection to %s[%p] closed (I/O condition=%d)", client->name, client, condition);
keep = FALSE;
} else if(condition & (G_IO_HUP|G_IO_NVAL|G_IO_ERR)) {
crm_trace("The connection %s[%p] has been closed (I/O condition=%d, refcount=%d)",
client->name, client, condition, mainloop_gio_refcount(client));
keep = FALSE;
} else if((condition & G_IO_IN) == 0) {
/*
#define GLIB_SYSDEF_POLLIN =1
#define GLIB_SYSDEF_POLLPRI =2
#define GLIB_SYSDEF_POLLOUT =4
#define GLIB_SYSDEF_POLLERR =8
#define GLIB_SYSDEF_POLLHUP =16
#define GLIB_SYSDEF_POLLNVAL =32
typedef enum
{
G_IO_IN GLIB_SYSDEF_POLLIN,
G_IO_OUT GLIB_SYSDEF_POLLOUT,
G_IO_PRI GLIB_SYSDEF_POLLPRI,
G_IO_ERR GLIB_SYSDEF_POLLERR,
G_IO_HUP GLIB_SYSDEF_POLLHUP,
G_IO_NVAL GLIB_SYSDEF_POLLNVAL
} GIOCondition;
A bitwise combination representing a condition to watch for on an event source.
G_IO_IN There is data to read.
G_IO_OUT Data can be written (without blocking).
G_IO_PRI There is urgent data to read.
G_IO_ERR Error condition.
G_IO_HUP Hung up (the connection has been broken, usually for pipes and sockets).
G_IO_NVAL Invalid request. The file descriptor is not open.
*/
crm_err("Strange condition: %d", condition);
}
/* keep == FALSE results in mainloop_gio_destroy() being called
* just before the source is removed from mainloop
*/
return keep;
}
static bool throttle_cib_load(float *load)
{
/*
/proc/[pid]/stat
Status information about the process. This is used by ps(1). It is defined in /usr/src/linux/fs/proc/array.c.
The fields, in order, with their proper scanf(3) format specifiers, are:
pid %d (1) The process ID.
comm %s (2) The filename of the executable, in parentheses. This is visible whether or not the executable is swapped out.
state %c (3) One character from the string "RSDZTW" where R is running, S is sleeping in an interruptible wait, D is waiting in uninterruptible disk sleep, Z is zombie, T is traced or stopped (on a signal), and W is paging.
ppid %d (4) The PID of the parent.
pgrp %d (5) The process group ID of the process.
session %d (6) The session ID of the process.
tty_nr %d (7) The controlling terminal of the process. (The minor device number is contained in the combination of bits 31 to 20 and 7 to 0; the major device number is in bits 15 to 8.)
tpgid %d (8) The ID of the foreground process group of the controlling terminal of the process.
flags %u (%lu before Linux 2.6.22)
(9) The kernel flags word of the process. For bit meanings, see the PF_* defines in the Linux kernel source file include/linux/sched.h. Details depend on the kernel version.
minflt %lu (10) The number of minor faults the process has made which have not required loading a memory page from disk.
cminflt %lu (11) The number of minor faults that the process's waited-for children have made.
majflt %lu (12) The number of major faults the process has made which have required loading a memory page from disk.
cmajflt %lu (13) The number of major faults that the process's waited-for children have made.
utime %lu (14) Amount of time that this process has been scheduled in user mode, measured in clock ticks (divide by sysconf(_SC_CLK_TCK)). This includes guest time, guest_time (time spent running a virtual CPU, see below), so that applications that are not aware of the guest time field do not lose that time from their calculations.
stime %lu (15) Amount of time that this process has been scheduled in kernel mode, measured in clock ticks (divide by sysconf(_SC_CLK_TCK)).
*/
static char *loadfile = NULL;
static time_t last_call = 0;
static long ticks_per_s = 0;
static unsigned long last_utime, last_stime;
char buffer[64*1024];
FILE *stream = NULL;
time_t now = time(NULL);
if(load == NULL) {
return FALSE;
} else {
*load = 0.0;
}
if(loadfile == NULL) {
last_call = 0;
last_utime = 0;
last_stime = 0;
loadfile = find_cib_loadfile();
ticks_per_s = sysconf(_SC_CLK_TCK);
crm_trace("Found %s", loadfile);
}
stream = fopen(loadfile, "r");
if(stream == NULL) {
int rc = errno;
crm_warn("Couldn't read %s: %s (%d)", loadfile, pcmk_strerror(rc), rc);
free(loadfile); loadfile = NULL;
return FALSE;
}
if(fgets(buffer, sizeof(buffer), stream)) {
char *comm = calloc(1, 256);
char state = 0;
int rc = 0, pid = 0, ppid = 0, pgrp = 0, session = 0, tty_nr = 0, tpgid = 0;
unsigned long flags = 0, minflt = 0, cminflt = 0, majflt = 0, cmajflt = 0, utime = 0, stime = 0;
rc = sscanf(buffer, "%d %[^ ] %c %d %d %d %d %d %lu %lu %lu %lu %lu %lu %lu",
&pid, comm, &state,
&ppid, &pgrp, &session, &tty_nr, &tpgid,
&flags, &minflt, &cminflt, &majflt, &cmajflt, &utime, &stime);
free(comm);
if(rc != 15) {
crm_err("Only %d of 15 fields found in %s", rc, loadfile);
fclose(stream);
return FALSE;
} else if(last_call > 0
&& last_call < now
&& last_utime <= utime
&& last_stime <= stime) {
time_t elapsed = now - last_call;
unsigned long delta_utime = utime - last_utime;
unsigned long delta_stime = stime - last_stime;
*load = (delta_utime + delta_stime); /* Cast to a float before division */
*load /= ticks_per_s;
*load /= elapsed;
crm_debug("cib load: %f (%lu ticks in %ds)", *load, delta_utime + delta_stime, elapsed);
} else {
crm_debug("Init %lu + %lu ticks at %d (%lu tps)", utime, stime, now, ticks_per_s);
}
last_call = now;
last_utime = utime;
last_stime = stime;
fclose(stream);
return TRUE;
}
fclose(stream);
return FALSE;
}
int
main(int argc, char *argv[])
{
int argerr = 0;
int flag;
int index = 0;
int rc = 0;
servicelog *slog = NULL;
struct sl_event *event = NULL;
uint64_t event_id = 0;
crm_log_cli_init("notifyServicelogEvent");
crm_set_options(NULL, "event_id ", long_options,
"Gets called upon events written to servicelog database");
if (argc < 2) {
argerr++;
}
while (1) {
flag = crm_get_option(argc, argv, &index);
if (flag == -1)
break;
switch (flag) {
case '?':
case '$':
crm_help(flag, CRM_EX_OK);
break;
default:
++argerr;
break;
}
}
if (argc - optind != 1) {
++argerr;
}
if (argerr) {
crm_help('?', CRM_EX_USAGE);
}
openlog("notifyServicelogEvent", LOG_NDELAY, LOG_USER);
if (sscanf(argv[optind], "%" U64TS, &event_id) != 1) {
crm_err("Error: could not read event_id from args!");
rc = 1;
goto cleanup;
}
if (event_id == 0) {
crm_err("Error: event_id is 0!");
rc = 1;
goto cleanup;
}
rc = servicelog_open(&slog, 0); /* flags is one of SL_FLAG_xxx */
if (!slog) {
crm_err("Error: servicelog_open failed, rc = %d", rc);
rc = 1;
goto cleanup;
}
if (slog) {
rc = servicelog_event_get(slog, event_id, &event);
}
if (rc == 0) {
STATUS status = STATUS_GREEN;
const char *health_component = "#health-ipmi";
const char *health_status = NULL;
crm_debug("Event id = %" U64T ", Log timestamp = %s, Event timestamp = %s",
event_id, ctime(&(event->time_logged)), ctime(&(event->time_event)));
status = event2status(event);
health_status = status2char(status);
if (health_status) {
gboolean rc;
/* @TODO pass attrd_opt_remote when appropriate */
rc = (attrd_update_delegate(NULL, 'v', NULL, health_component,
health_status, NULL, NULL, NULL, NULL,
attrd_opt_none) > 0);
crm_debug("Updating attribute ('%s', '%s') = %d",
health_component, health_status, rc);
} else {
crm_err("Error: status2char failed, status = %d", status);
rc = 1;
}
} else {
crm_err("Error: servicelog_event_get failed, rc = %d", rc);
}
cleanup:
if (event) {
servicelog_event_free(event);
}
if (slog) {
servicelog_close(slog);
}
closelog();
return rc;
}
int
register_fsa_input_adv(enum crmd_fsa_cause cause, enum crmd_fsa_input input,
void *data, long long with_actions,
gboolean prepend, const char *raised_from)
{
unsigned old_len = g_list_length(fsa_message_queue);
fsa_data_t *fsa_data = NULL;
last_data_id++;
CRM_CHECK(raised_from != NULL, raised_from = "<unknown>");
crm_trace("%s %s FSA input %d (%s) (cause=%s) %s data",
raised_from, prepend ? "prepended" : "appended", last_data_id,
fsa_input2string(input), fsa_cause2string(cause), data ? "with" : "without");
if (input == I_WAIT_FOR_EVENT) {
do_fsa_stall = TRUE;
crm_debug("Stalling the FSA pending further input: cause=%s", fsa_cause2string(cause));
if (old_len > 0) {
crm_warn("%s stalled the FSA with pending inputs", raised_from);
fsa_dump_queue(LOG_DEBUG);
}
if (data == NULL) {
set_bit_inplace(fsa_actions, with_actions);
with_actions = A_NOTHING;
return 0;
}
crm_err("%s stalled the FSA with data - this may be broken", raised_from);
}
if (input == I_NULL && with_actions == A_NOTHING /* && data == NULL */ ) {
/* no point doing anything */
crm_err("Cannot add entry to queue: no input and no action");
return 0;
}
crm_malloc0(fsa_data, sizeof(fsa_data_t));
fsa_data->id = last_data_id;
fsa_data->fsa_input = input;
fsa_data->fsa_cause = cause;
fsa_data->origin = raised_from;
fsa_data->data = NULL;
fsa_data->data_type = fsa_dt_none;
fsa_data->actions = with_actions;
if (with_actions != A_NOTHING) {
crm_trace("Adding actions %.16llx to input", with_actions);
}
if (data != NULL) {
switch (cause) {
case C_FSA_INTERNAL:
case C_CRMD_STATUS_CALLBACK:
case C_IPC_MESSAGE:
case C_HA_MESSAGE:
crm_trace("Copying %s data from %s as a HA msg",
fsa_cause2string(cause), raised_from);
CRM_CHECK(((ha_msg_input_t *) data)->msg != NULL,
crm_err("Bogus data from %s", raised_from));
fsa_data->data = copy_ha_msg_input(data);
fsa_data->data_type = fsa_dt_ha_msg;
break;
case C_LRM_OP_CALLBACK:
crm_trace("Copying %s data from %s as lrm_op_t",
fsa_cause2string(cause), raised_from);
fsa_data->data = copy_lrm_op((lrm_op_t *) data);
fsa_data->data_type = fsa_dt_lrm;
break;
case C_CCM_CALLBACK:
case C_SUBSYSTEM_CONNECT:
case C_LRM_MONITOR_CALLBACK:
case C_TIMER_POPPED:
case C_SHUTDOWN:
case C_HEARTBEAT_FAILED:
case C_HA_DISCONNECT:
case C_ILLEGAL:
case C_UNKNOWN:
case C_STARTUP:
crm_err("Copying %s data (from %s)"
" not yet implemented", fsa_cause2string(cause), raised_from);
exit(1);
break;
}
crm_trace("%s data copied", fsa_cause2string(fsa_data->fsa_cause));
}
/* make sure to free it properly later */
if (prepend) {
crm_trace("Prepending input");
fsa_message_queue = g_list_prepend(fsa_message_queue, fsa_data);
} else {
fsa_message_queue = g_list_append(fsa_message_queue, fsa_data);
}
crm_trace("Queue len: %d", g_list_length(fsa_message_queue));
fsa_dump_queue(LOG_DEBUG_2);
if (old_len == g_list_length(fsa_message_queue)) {
crm_err("Couldnt add message to the queue");
}
if (fsa_source) {
crm_trace("Triggering FSA: %s", __FUNCTION__);
mainloop_set_trigger(fsa_source);
}
return last_data_id;
}
enum crmd_fsa_input
handle_request(xmlNode * stored_msg)
{
xmlNode *msg = NULL;
const char *op = crm_element_value(stored_msg, F_CRM_TASK);
/* Optimize this for the DC - it has the most to do */
if (op == NULL) {
crm_log_xml_err(stored_msg, "Bad message");
return I_NULL;
}
/*========== DC-Only Actions ==========*/
if (AM_I_DC) {
if (strcmp(op, CRM_OP_JOIN_ANNOUNCE) == 0) {
return I_NODE_JOIN;
} else if (strcmp(op, CRM_OP_JOIN_REQUEST) == 0) {
return I_JOIN_REQUEST;
} else if (strcmp(op, CRM_OP_JOIN_CONFIRM) == 0) {
return I_JOIN_RESULT;
} else if (strcmp(op, CRM_OP_SHUTDOWN) == 0) {
const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM);
gboolean dc_match = safe_str_eq(host_from, fsa_our_dc);
if (is_set(fsa_input_register, R_SHUTDOWN)) {
crm_info("Shutting ourselves down (DC)");
return I_STOP;
} else if (dc_match) {
crm_err("We didnt ask to be shut down, yet our"
" TE is telling us too." " Better get out now!");
return I_TERMINATE;
} else if (fsa_state != S_STOPPING) {
crm_err("Another node is asking us to shutdown" " but we think we're ok.");
return I_ELECTION;
}
} else if (strcmp(op, CRM_OP_SHUTDOWN_REQ) == 0) {
/* a slave wants to shut down */
/* create cib fragment and add to message */
return handle_shutdown_request(stored_msg);
}
}
/*========== common actions ==========*/
if (strcmp(op, CRM_OP_NOVOTE) == 0) {
ha_msg_input_t fsa_input;
fsa_input.msg = stored_msg;
register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input,
A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __FUNCTION__);
} else if (strcmp(op, CRM_OP_CLEAR_FAILCOUNT) == 0) {
return handle_failcount_op(stored_msg);
} else if (strcmp(op, CRM_OP_VOTE) == 0) {
/* count the vote and decide what to do after that */
ha_msg_input_t fsa_input;
fsa_input.msg = stored_msg;
register_fsa_input_adv(C_HA_MESSAGE, I_NULL, &fsa_input,
A_ELECTION_COUNT | A_ELECTION_CHECK, FALSE, __FUNCTION__);
/* Sometimes we _must_ go into S_ELECTION */
if (fsa_state == S_HALT) {
crm_debug("Forcing an election from S_HALT");
return I_ELECTION;
#if 0
} else if (AM_I_DC) {
/* This is the old way of doing things but what is gained? */
return I_ELECTION;
#endif
}
} else if (strcmp(op, CRM_OP_JOIN_OFFER) == 0) {
crm_debug("Raising I_JOIN_OFFER: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID));
return I_JOIN_OFFER;
} else if (strcmp(op, CRM_OP_JOIN_ACKNAK) == 0) {
crm_debug("Raising I_JOIN_RESULT: join-%s", crm_element_value(stored_msg, F_CRM_JOIN_ID));
return I_JOIN_RESULT;
} else if (strcmp(op, CRM_OP_LRM_DELETE) == 0
|| strcmp(op, CRM_OP_LRM_FAIL) == 0
|| strcmp(op, CRM_OP_LRM_REFRESH) == 0 || strcmp(op, CRM_OP_REPROBE) == 0) {
crm_xml_add(stored_msg, F_CRM_SYS_TO, CRM_SYSTEM_LRMD);
return I_ROUTER;
} else if (strcmp(op, CRM_OP_NOOP) == 0) {
return I_NULL;
} else if (strcmp(op, CRM_OP_LOCAL_SHUTDOWN) == 0) {
crm_shutdown(SIGTERM);
/*return I_SHUTDOWN; */
return I_NULL;
/*========== (NOT_DC)-Only Actions ==========*/
} else if (AM_I_DC == FALSE && strcmp(op, CRM_OP_SHUTDOWN) == 0) {
const char *host_from = crm_element_value(stored_msg, F_CRM_HOST_FROM);
gboolean dc_match = safe_str_eq(host_from, fsa_our_dc);
if (dc_match || fsa_our_dc == NULL) {
if (is_set(fsa_input_register, R_SHUTDOWN) == FALSE) {
crm_err("We didn't ask to be shut down, yet our" " DC is telling us too.");
set_bit_inplace(fsa_input_register, R_STAYDOWN);
return I_STOP;
}
crm_info("Shutting down");
return I_STOP;
} else {
crm_warn("Discarding %s op from %s", op, host_from);
}
} else if (strcmp(op, CRM_OP_PING) == 0) {
/* eventually do some stuff to figure out
* if we /are/ ok
*/
const char *sys_to = crm_element_value(stored_msg, F_CRM_SYS_TO);
xmlNode *ping = createPingAnswerFragment(sys_to, "ok");
crm_xml_add(ping, "crmd_state", fsa_state2string(fsa_state));
crm_info("Current ping state: %s", fsa_state2string(fsa_state));
msg = create_reply(stored_msg, ping);
relay_message(msg, TRUE);
free_xml(ping);
free_xml(msg);
/* probably better to do this via signals on the
* local node
*/
} else if (strcmp(op, CRM_OP_DEBUG_UP) == 0) {
crm_bump_log_level();
crm_info("Debug set to %d", get_crm_log_level());
} else if (strcmp(op, CRM_OP_DEBUG_DOWN) == 0) {
alter_debug(DEBUG_DEC);
crm_info("Debug set to %d", get_crm_log_level());
} else {
crm_err("Unexpected request (%s) sent to %s", op, AM_I_DC ? "the DC" : "non-DC node");
crm_log_xml_err(stored_msg, "Unexpected");
}
return I_NULL;
}
gboolean
mcp_read_config(void)
{
int rc = CS_OK;
int retries = 0;
const char *const_value = NULL;
#if HAVE_CONFDB
char *value = NULL;
confdb_handle_t config = 0;
confdb_handle_t top_handle = 0;
hdb_handle_t local_handle;
static confdb_callbacks_t callbacks = { };
do {
rc = confdb_initialize(&config, &callbacks);
if (rc != CS_OK) {
retries++;
printf("confdb connection setup failed: %s. Retrying in %ds\n", ais_error2text(rc), retries);
crm_info("confdb connection setup failed: %s. Retrying in %ds", ais_error2text(rc), retries);
sleep(retries);
} else {
break;
}
} while (retries < 5);
#elif HAVE_CMAP
cmap_handle_t local_handle;
uint64_t config = 0;
/* There can be only one (possibility if confdb isn't around) */
do {
rc = cmap_initialize(&local_handle);
if (rc != CS_OK) {
retries++;
printf("cmap connection setup failed: %s. Retrying in %ds\n", cs_strerror(rc), retries);
crm_info("cmap connection setup failed: %s. Retrying in %ds", cs_strerror(rc), retries);
sleep(retries);
} else {
break;
}
} while (retries < 5);
#endif
if (rc != CS_OK) {
printf("Could not connect to Cluster Configuration Database API, error %d\n", rc);
crm_warn("Could not connect to Cluster Configuration Database API, error %d", rc);
return FALSE;
}
stack = get_cluster_type();
crm_info("Reading configure for stack: %s", name_for_cluster_type(stack));
/* =::=::= Should we be here =::=::= */
if (stack == pcmk_cluster_corosync) {
set_daemon_option("cluster_type", "corosync");
set_daemon_option("quorum_type", "corosync");
#if HAVE_CONFDB
} else if (stack == pcmk_cluster_cman) {
set_daemon_option("cluster_type", "cman");
set_daemon_option("quorum_type", "cman");
enable_crmd_as_root(TRUE);
} else if (stack == pcmk_cluster_classic_ais) {
set_daemon_option("cluster_type", "openais");
set_daemon_option("quorum_type", "pcmk");
/* Look for a service block to indicate our plugin is loaded */
top_handle = config_find_init(config);
local_handle = config_find_next(config, "service", top_handle);
while (local_handle) {
get_config_opt(config, local_handle, "name", &value, NULL);
if (safe_str_eq("pacemaker", value)) {
get_config_opt(config, local_handle, "ver", &value, "0");
if (safe_str_eq(value, "1")) {
get_config_opt(config, local_handle, "use_logd", &value, "no");
set_daemon_option("use_logd", value);
set_daemon_option("LOGD", value);
get_config_opt(config, local_handle, "use_mgmtd", &value, "no");
enable_mgmtd(crm_is_true(value));
} else {
crm_err("We can only start Pacemaker from init if using version 1"
" of the Pacemaker plugin for Corosync. Terminating.");
crm_exit(DAEMON_RESPAWN_STOP);
}
break;
}
local_handle = config_find_next(config, "service", top_handle);
}
free(value);
#endif
} else {
crm_err("Unsupported stack type: %s", name_for_cluster_type(stack));
return FALSE;
}
#if HAVE_CONFDB
top_handle = config_find_init(config);
local_handle = config_find_next(config, "logging", top_handle);
#endif
/* =::=::= Logging =::=::= */
if (daemon_option("debug")) {
/* Syslog logging is already setup by crm_log_init() */
} else {
/* Check corosync */
char *debug_enabled = NULL;
get_config_opt(config, local_handle, KEY_PREFIX "debug", &debug_enabled, "off");
if (crm_is_true(debug_enabled)) {
set_daemon_option("debug", "1");
if (get_crm_log_level() < LOG_DEBUG) {
set_crm_log_level(LOG_DEBUG);
}
} else {
set_daemon_option("debug", "0");
}
free(debug_enabled);
}
const_value = daemon_option("debugfile");
if (daemon_option("logfile")) {
/* File logging is already setup by crm_log_init() */
} else if(const_value) {
/* From when we cared what options heartbeat used */
set_daemon_option("logfile", const_value);
crm_add_logfile(const_value);
} else {
/* Check corosync */
char *logfile = NULL;
char *logfile_enabled = NULL;
get_config_opt(config, local_handle, KEY_PREFIX "to_logfile", &logfile_enabled, "on");
get_config_opt(config, local_handle, KEY_PREFIX "logfile", &logfile, "/var/log/pacemaker.log");
if (crm_is_true(logfile_enabled) == FALSE) {
crm_trace("File logging disabled in corosync");
} else if (crm_add_logfile(logfile)) {
set_daemon_option("logfile", logfile);
} else {
crm_err("Couldn't create logfile: %s", logfile);
set_daemon_option("logfile", "none");
}
free(logfile);
free(logfile_enabled);
}
if (daemon_option("logfacility")) {
/* Syslog logging is already setup by crm_log_init() */
} else {
/* Check corosync */
char *syslog_enabled = NULL;
char *syslog_facility = NULL;
get_config_opt(config, local_handle, KEY_PREFIX "to_syslog", &syslog_enabled, "on");
get_config_opt(config, local_handle, KEY_PREFIX "syslog_facility", &syslog_facility, "daemon");
if (crm_is_true(syslog_enabled) == FALSE) {
qb_log_ctl(QB_LOG_SYSLOG, QB_LOG_CONF_ENABLED, QB_FALSE);
set_daemon_option("logfacility", "none");
} else {
qb_log_ctl(QB_LOG_SYSLOG, QB_LOG_CONF_FACILITY, qb_log_facility2int(syslog_facility));
qb_log_ctl(QB_LOG_SYSLOG, QB_LOG_CONF_ENABLED, QB_TRUE);
set_daemon_option("logfacility", syslog_facility);
}
free(syslog_enabled);
free(syslog_facility);
}
const_value = daemon_option("logfacility");
if (const_value) {
/* cluster-glue module needs HA_LOGFACILITY */
setenv("HA_LOGFACILITY", const_value, 1);
}
#if HAVE_CONFDB
confdb_finalize(config);
#elif HAVE_CMAP
if(local_handle){
gid_t gid = 0;
if (crm_user_lookup(CRM_DAEMON_USER, NULL, &gid) < 0) {
crm_warn("Could not authorize group with corosync " CRM_XS
" No group found for user %s", CRM_DAEMON_USER);
} else {
char key[PATH_MAX];
snprintf(key, PATH_MAX, "uidgid.gid.%u", gid);
rc = cmap_set_uint8(local_handle, key, 1);
if (rc != CS_OK) {
crm_warn("Could not authorize group with corosync "CRM_XS
" group=%u rc=%d (%s)", gid, rc, ais_error2text(rc));
}
}
}
cmap_finalize(local_handle);
#endif
return TRUE;
}
gboolean
upstart_job_exec(svc_action_t * op, gboolean synchronous)
{
char *job = NULL;
int arg_wait = TRUE;
const char *arg_env = "pacemaker=1";
const char *action = op->action;
DBusError error;
DBusMessage *msg = NULL;
DBusMessage *reply = NULL;
DBusMessageIter iter, array_iter;
op->rc = PCMK_OCF_UNKNOWN_ERROR;
CRM_ASSERT(upstart_init());
if (safe_str_eq(op->action, "meta-data")) {
op->stdout_data = upstart_job_metadata(op->agent);
op->rc = PCMK_OCF_OK;
goto cleanup;
}
if(!upstart_job_by_name(op->agent, &job)) {
crm_debug("Could not obtain job named '%s' to %s", op->agent, action);
if (!g_strcmp0(action, "stop")) {
op->rc = PCMK_OCF_OK;
} else {
op->rc = PCMK_OCF_NOT_INSTALLED;
op->status = PCMK_LRM_OP_NOT_INSTALLED;
}
goto cleanup;
}
if (safe_str_eq(op->action, "monitor") || safe_str_eq(action, "status")) {
if (upstart_job_running(op->agent)) {
op->rc = PCMK_OCF_OK;
} else {
op->rc = PCMK_OCF_NOT_RUNNING;
}
goto cleanup;
} else if (!g_strcmp0(action, "start")) {
action = "Start";
} else if (!g_strcmp0(action, "stop")) {
action = "Stop";
} else if (!g_strcmp0(action, "restart")) {
action = "Restart";
} else {
op->rc = PCMK_OCF_UNIMPLEMENT_FEATURE;
goto cleanup;
}
crm_debug("Calling %s for %s on %s", action, op->rsc, job);
msg = dbus_message_new_method_call(BUS_NAME, // target for the method call
job, // object to call on
UPSTART_JOB_IFACE, // interface to call on
action); // method name
CRM_ASSERT(msg != NULL);
dbus_message_iter_init_append (msg, &iter);
CRM_LOG_ASSERT(dbus_message_iter_open_container (&iter,
DBUS_TYPE_ARRAY,
DBUS_TYPE_STRING_AS_STRING,
&array_iter));
CRM_LOG_ASSERT(dbus_message_iter_append_basic (&array_iter, DBUS_TYPE_STRING, &arg_env));
CRM_LOG_ASSERT(dbus_message_iter_close_container (&iter, &array_iter));
CRM_LOG_ASSERT(dbus_message_append_args(msg, DBUS_TYPE_BOOLEAN, &arg_wait, DBUS_TYPE_INVALID));
if (synchronous == FALSE) {
free(job);
return pcmk_dbus_send(msg, upstart_proxy, upstart_async_dispatch, op);
}
dbus_error_init(&error);
reply = pcmk_dbus_send_recv(msg, upstart_proxy, &error);
if(error.name) {
if(!upstart_mask_error(op, error.name)) {
crm_err("Could not issue %s for %s: %s (%s)", action, op->rsc, error.name, job);
}
} else if (!g_strcmp0(op->action, "stop")) {
/* No return vaue */
op->rc = PCMK_OCF_OK;
} else if(!pcmk_dbus_type_check(reply, NULL, DBUS_TYPE_OBJECT_PATH, __FUNCTION__, __LINE__)) {
crm_warn("Call to %s passed but return type was unexpected", op->action);
op->rc = PCMK_OCF_OK;
} else {
const char *path = NULL;
dbus_message_get_args (reply, NULL,
DBUS_TYPE_OBJECT_PATH, &path,
DBUS_TYPE_INVALID);
crm_info("Call to %s passed: %s", op->action, path);
op->rc = PCMK_OCF_OK;
}
cleanup:
free(job);
if(msg) {
dbus_message_unref(msg);
}
if(reply) {
dbus_message_unref(reply);
}
if (synchronous == FALSE) {
operation_finalize(op);
return TRUE;
}
return op->rc == PCMK_OCF_OK;
}
int
main(int argc, char **argv)
{
int flag;
int argerr = 0;
gboolean allow_cores = TRUE;
IPC_Channel *old_instance = NULL;
crm_system_name = CRM_SYSTEM_PENGINE;
mainloop_add_signal(SIGTERM, pengine_shutdown);
while ((flag = getopt(argc, argv, OPTARGS)) != EOF) {
switch (flag) {
case 'V':
alter_debug(DEBUG_INC);
break;
case 'h': /* Help message */
usage(crm_system_name, LSB_EXIT_OK);
break;
case 'c':
allow_cores = TRUE;
break;
default:
++argerr;
break;
}
}
if (argc - optind == 1 && safe_str_eq("metadata", argv[optind])) {
pe_metadata();
return 0;
}
if (optind > argc) {
++argerr;
}
if (argerr) {
usage(crm_system_name, LSB_EXIT_GENERIC);
}
crm_log_init(NULL, LOG_NOTICE, TRUE, FALSE, argc, argv);
if (crm_is_writable(PE_STATE_DIR, NULL, CRM_DAEMON_USER, CRM_DAEMON_GROUP, FALSE) == FALSE) {
crm_err("Bad permissions on " PE_STATE_DIR ". Terminating");
fprintf(stderr, "ERROR: Bad permissions on " PE_STATE_DIR ". See logs for details\n");
fflush(stderr);
return 100;
}
ipc_server = crm_strdup(CRM_SYSTEM_PENGINE);
/* find any previous instances and shut them down */
crm_debug("Checking for old instances of %s", crm_system_name);
old_instance = init_client_ipc_comms_nodispatch(CRM_SYSTEM_PENGINE);
while (old_instance != NULL) {
xmlNode *cmd =
create_request(CRM_OP_QUIT, NULL, NULL, CRM_SYSTEM_PENGINE, CRM_SYSTEM_PENGINE, NULL);
crm_warn("Terminating previous PE instance");
send_ipc_message(old_instance, cmd);
free_xml(cmd);
sleep(2);
old_instance->ops->destroy(old_instance);
old_instance = init_client_ipc_comms_nodispatch(CRM_SYSTEM_PENGINE);
}
crm_debug("Init server comms");
if (init_server_ipc_comms(ipc_server, pe_client_connect, default_ipc_connection_destroy)) {
crm_err("Couldn't start IPC server");
return 1;
}
/* Create the mainloop and run it... */
crm_info("Starting %s", crm_system_name);
mainloop = g_main_new(FALSE);
g_main_run(mainloop);
#if HAVE_LIBXML2
crm_xml_cleanup();
#endif
crm_info("Exiting %s", crm_system_name);
return 0;
}
const char *
pcmk_strerror(int rc)
{
int error = abs(rc);
if (error == 0) {
return "OK";
} else if (error < PCMK_ERROR_OFFSET) {
return strerror(error);
}
switch (error) {
case pcmk_err_generic:
return "Generic Pacemaker error";
case pcmk_err_no_quorum:
return "Operation requires quorum";
case pcmk_err_schema_validation:
return "Update does not conform to the configured schema";
case pcmk_err_transform_failed:
return "Schema transform failed";
case pcmk_err_old_data:
return "Update was older than existing configuration";
case pcmk_err_diff_failed:
return "Application of an update diff failed";
case pcmk_err_diff_resync:
return "Application of an update diff failed, requesting a full refresh";
case pcmk_err_cib_modified:
return "The on-disk configuration was manually modified";
case pcmk_err_cib_backup:
return "Could not archive the previous configuration";
case pcmk_err_cib_save:
return "Could not save the new configuration to disk";
case pcmk_err_cib_corrupt:
return "Could not parse on-disk configuration";
case pcmk_err_multiple:
return "Resource active on multiple nodes";
case pcmk_err_node_unknown:
return "Node not found";
case pcmk_err_already:
return "Situation already as requested";
case pcmk_err_bad_nvpair:
return "Bad name/value pair given";
case pcmk_err_schema_unchanged:
return "Schema is already the latest available";
/* The following cases will only be hit on systems for which they are non-standard */
/* coverity[dead_error_condition] False positive on non-Linux */
case ENOTUNIQ:
return "Name not unique on network";
/* coverity[dead_error_condition] False positive on non-Linux */
case ECOMM:
return "Communication error on send";
/* coverity[dead_error_condition] False positive on non-Linux */
case ELIBACC:
return "Can not access a needed shared library";
/* coverity[dead_error_condition] False positive on non-Linux */
case EREMOTEIO:
return "Remote I/O error";
/* coverity[dead_error_condition] False positive on non-Linux */
case EUNATCH:
return "Protocol driver not attached";
/* coverity[dead_error_condition] False positive on non-Linux */
case ENOKEY:
return "Required key not available";
}
crm_err("Unknown error code: %d", rc);
return "Unknown error";
}
static void
action_launch_child(svc_action_t *op)
{
int lpc;
/* SIGPIPE is ignored (which is different from signal blocking) by the gnutls library.
* Depending on the libqb version in use, libqb may set SIGPIPE to be ignored as well.
* We do not want this to be inherited by the child process. By resetting this the signal
* to the default behavior, we avoid some potential odd problems that occur during OCF
* scripts when SIGPIPE is ignored by the environment. */
signal(SIGPIPE, SIG_DFL);
#if defined(HAVE_SCHED_SETSCHEDULER)
if (sched_getscheduler(0) != SCHED_OTHER) {
struct sched_param sp;
memset(&sp, 0, sizeof(sp));
sp.sched_priority = 0;
if (sched_setscheduler(0, SCHED_OTHER, &sp) == -1) {
crm_perror(LOG_ERR, "Could not reset scheduling policy to SCHED_OTHER for %s", op->id);
}
}
#endif
if (setpriority(PRIO_PROCESS, 0, 0) == -1) {
crm_perror(LOG_ERR, "Could not reset process priority to 0 for %s", op->id);
}
/* Man: The call setpgrp() is equivalent to setpgid(0,0)
* _and_ compiles on BSD variants too
* need to investigate if it works the same too.
*/
setpgid(0, 0);
/* close all descriptors except stdin/out/err and channels to logd */
for (lpc = getdtablesize() - 1; lpc > STDERR_FILENO; lpc--) {
close(lpc);
}
#if SUPPORT_CIBSECRETS
if (replace_secret_params(op->rsc, op->params) < 0) {
/* replacing secrets failed! */
if (safe_str_eq(op->action,"stop")) {
/* don't fail on stop! */
crm_info("proceeding with the stop operation for %s", op->rsc);
} else {
crm_err("failed to get secrets for %s, "
"considering resource not configured", op->rsc);
_exit(PCMK_OCF_NOT_CONFIGURED);
}
}
#endif
/* Setup environment correctly */
add_OCF_env_vars(op);
/* execute the RA */
execvp(op->opaque->exec, op->opaque->args);
/* Most cases should have been already handled by stat() */
services_handle_exec_error(op, errno);
_exit(op->rc);
}
void
tengine_stonith_callback(stonith_t * stonith, stonith_callback_data_t * data)
{
char *uuid = NULL;
int stonith_id = -1;
int transition_id = -1;
crm_action_t *action = NULL;
int call_id = data->call_id;
int rc = data->rc;
char *userdata = data->userdata;
CRM_CHECK(userdata != NULL, return);
crm_notice("Stonith operation %d/%s: %s (%d)", call_id, (char *)userdata,
pcmk_strerror(rc), rc);
if (AM_I_DC == FALSE) {
return;
}
/* crm_info("call=%d, optype=%d, node_name=%s, result=%d, node_list=%s, action=%s", */
/* op->call_id, op->optype, op->node_name, op->op_result, */
/* (char *)op->node_list, op->private_data); */
/* filter out old STONITH actions */
CRM_CHECK(decode_transition_key(userdata, &uuid, &transition_id, &stonith_id, NULL),
goto bail);
if (transition_graph->complete || stonith_id < 0 || safe_str_neq(uuid, te_uuid)
|| transition_graph->id != transition_id) {
crm_info("Ignoring STONITH action initiated outside of the current transition");
goto bail;
}
action = get_action(stonith_id, FALSE);
if (action == NULL) {
crm_err("Stonith action not matched");
goto bail;
}
stop_te_timer(action->timer);
if (rc == pcmk_ok) {
const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET);
const char *uuid = crm_element_value(action->xml, XML_LRM_ATTR_TARGET_UUID);
const char *op = crm_meta_value(action->params, "stonith_action");
crm_info("Stonith operation %d for %s passed", call_id, target);
if (action->confirmed == FALSE) {
te_action_confirmed(action);
if (safe_str_eq("on", op)) {
const char *value = NULL;
char *now = crm_itoa(time(NULL));
update_attrd(target, CRM_ATTR_UNFENCED, now, NULL, FALSE);
free(now);
value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_ALL);
update_attrd(target, CRM_ATTR_DIGESTS_ALL, value, NULL, FALSE);
value = crm_meta_value(action->params, XML_OP_ATTR_DIGESTS_SECURE);
update_attrd(target, CRM_ATTR_DIGESTS_SECURE, value, NULL, FALSE);
} else if (action->sent_update == FALSE) {
send_stonith_update(action, target, uuid);
action->sent_update = TRUE;
}
}
st_fail_count_reset(target);
} else {
const char *target = crm_element_value(action->xml, XML_LRM_ATTR_TARGET);
enum transition_action abort_action = tg_restart;
action->failed = TRUE;
crm_notice("Stonith operation %d for %s failed (%s): aborting transition.",
call_id, target, pcmk_strerror(rc));
/* If no fence devices were available, there's no use in immediately
* checking again, so don't start a new transition in that case.
*/
if (rc == -ENODEV) {
crm_warn("No devices found in cluster to fence %s, giving up",
target);
abort_action = tg_stop;
}
/* Increment the fail count now, so abort_for_stonith_failure() can
* check it. Non-DC nodes will increment it in tengine_stonith_notify().
*/
st_fail_count_increment(target);
abort_for_stonith_failure(abort_action, target, NULL);
}
update_graph(transition_graph, action);
trigger_graph();
bail:
free(userdata);
free(uuid);
return;
}
/* Returns FALSE if 'op' should be free'd by the caller */
gboolean
services_os_action_execute(svc_action_t * op, gboolean synchronous)
{
int stdout_fd[2];
int stderr_fd[2];
sigset_t mask;
sigset_t old_mask;
struct stat st;
if (pipe(stdout_fd) < 0) {
crm_err("pipe() failed");
}
if (pipe(stderr_fd) < 0) {
crm_err("pipe() failed");
}
/* Fail fast */
if(stat(op->opaque->exec, &st) != 0) {
int rc = errno;
crm_warn("Cannot execute '%s': %s (%d)", op->opaque->exec, pcmk_strerror(rc), rc);
services_handle_exec_error(op, rc);
if (!synchronous) {
return operation_finalize(op);
}
return FALSE;
}
if (synchronous) {
sigemptyset(&mask);
sigaddset(&mask, SIGCHLD);
sigemptyset(&old_mask);
if (sigprocmask(SIG_BLOCK, &mask, &old_mask) < 0) {
crm_perror(LOG_ERR, "sigprocmask() failed");
}
}
op->pid = fork();
switch (op->pid) {
case -1:
{
int rc = errno;
close(stdout_fd[0]);
close(stdout_fd[1]);
close(stderr_fd[0]);
close(stderr_fd[1]);
crm_err("Could not execute '%s': %s (%d)", op->opaque->exec, pcmk_strerror(rc), rc);
services_handle_exec_error(op, rc);
if (!synchronous) {
return operation_finalize(op);
}
return FALSE;
}
case 0: /* Child */
close(stdout_fd[0]);
close(stderr_fd[0]);
if (STDOUT_FILENO != stdout_fd[1]) {
if (dup2(stdout_fd[1], STDOUT_FILENO) != STDOUT_FILENO) {
crm_err("dup2() failed (stdout)");
}
close(stdout_fd[1]);
}
if (STDERR_FILENO != stderr_fd[1]) {
if (dup2(stderr_fd[1], STDERR_FILENO) != STDERR_FILENO) {
crm_err("dup2() failed (stderr)");
}
close(stderr_fd[1]);
}
action_launch_child(op);
}
/* Only the parent reaches here */
close(stdout_fd[1]);
close(stderr_fd[1]);
op->opaque->stdout_fd = stdout_fd[0];
set_fd_opts(op->opaque->stdout_fd, O_NONBLOCK);
op->opaque->stderr_fd = stderr_fd[0];
set_fd_opts(op->opaque->stderr_fd, O_NONBLOCK);
if (synchronous) {
action_synced_wait(op, mask);
if (sigismember(&old_mask, SIGCHLD) == 0) {
if (sigprocmask(SIG_UNBLOCK, &mask, NULL) < 0) {
crm_perror(LOG_ERR, "sigprocmask() to unblocked failed");
}
}
} else {
crm_trace("Async waiting for %d - %s", op->pid, op->opaque->exec);
mainloop_child_add_with_flags(op->pid,
op->timeout,
op->id,
op,
(op->flags & SVC_ACTION_LEAVE_GROUP) ? mainloop_leave_pid_group : 0,
operation_finished);
op->opaque->stdout_gsource = mainloop_add_fd(op->id,
G_PRIORITY_LOW,
op->opaque->stdout_fd, op, &stdout_callbacks);
op->opaque->stderr_gsource = mainloop_add_fd(op->id,
G_PRIORITY_LOW,
op->opaque->stderr_fd, op, &stderr_callbacks);
}
return TRUE;
}
gboolean
read_config(void)
{
int rc = CS_OK;
int retries = 0;
gboolean have_log = FALSE;
const char *const_value = NULL;
char *logging_debug = NULL;
char *logging_logfile = NULL;
char *logging_to_logfile = NULL;
char *logging_to_syslog = NULL;
char *logging_syslog_facility = NULL;
#if HAVE_CONFDB
char *value = NULL;
confdb_handle_t config = 0;
confdb_handle_t top_handle = 0;
hdb_handle_t local_handle;
static confdb_callbacks_t callbacks = { };
do {
rc = confdb_initialize(&config, &callbacks);
if (rc != CS_OK) {
retries++;
printf("confdb connection setup failed: %s. Retrying in %ds\n", ais_error2text(rc), retries);
crm_info("confdb connection setup failed: %s. Retrying in %ds", ais_error2text(rc), retries);
sleep(retries);
} else {
break;
}
} while (retries < 5);
#elif HAVE_CMAP
cmap_handle_t local_handle;
uint64_t config = 0;
/* There can be only one (possibility if confdb isn't around) */
do {
rc = cmap_initialize(&local_handle);
if (rc != CS_OK) {
retries++;
printf("cmap connection setup failed: %s. Retrying in %ds\n", cs_strerror(rc), retries);
crm_info("cmap connection setup failed: %s. Retrying in %ds", cs_strerror(rc), retries);
sleep(retries);
} else {
break;
}
} while (retries < 5);
#endif
if (rc != CS_OK) {
printf("Could not connect to Cluster Configuration Database API, error %d\n", rc);
crm_warn("Could not connect to Cluster Configuration Database API, error %d", rc);
return FALSE;
}
stack = get_cluster_type();
crm_info("Reading configure for stack: %s", name_for_cluster_type(stack));
/* =::=::= Should we be here =::=::= */
if (stack == pcmk_cluster_corosync) {
set_daemon_option("cluster_type", "corosync");
set_daemon_option("quorum_type", "corosync");
#if HAVE_CONFDB
} else if (stack == pcmk_cluster_cman) {
set_daemon_option("cluster_type", "cman");
set_daemon_option("quorum_type", "cman");
enable_crmd_as_root(TRUE);
} else if (stack == pcmk_cluster_classic_ais) {
set_daemon_option("cluster_type", "openais");
set_daemon_option("quorum_type", "pcmk");
/* Look for a service block to indicate our plugin is loaded */
top_handle = config_find_init(config);
local_handle = config_find_next(config, "service", top_handle);
while (local_handle) {
get_config_opt(config, local_handle, "name", &value, NULL);
if (safe_str_eq("pacemaker", value)) {
get_config_opt(config, local_handle, "ver", &value, "0");
if (safe_str_eq(value, "1")) {
get_config_opt(config, local_handle, "use_logd", &value, "no");
set_daemon_option("use_logd", value);
set_daemon_option("LOGD", value);
get_config_opt(config, local_handle, "use_mgmtd", &value, "no");
enable_mgmtd(crm_is_true(value));
} else {
crm_err("We can only start Pacemaker from init if using version 1"
" of the Pacemaker plugin for Corosync. Terminating.");
crm_exit(DAEMON_RESPAWN_STOP);
}
break;
}
local_handle = config_find_next(config, "service", top_handle);
}
free(value);
#endif
} else {
crm_err("Unsupported stack type: %s", name_for_cluster_type(stack));
return FALSE;
}
#if HAVE_CONFDB
top_handle = config_find_init(config);
local_handle = config_find_next(config, "logging", top_handle);
#endif
/* =::=::= Logging =::=::= */
get_config_opt(config, local_handle, KEY_PREFIX "debug", &logging_debug, "off");
const_value = daemon_option("debugfile");
if (const_value) {
logging_to_logfile = strdup("on");
logging_logfile = strdup(const_value);
crm_trace("Using debugfile setting from the environment: %s", logging_logfile);
} else {
get_config_opt(config, local_handle, KEY_PREFIX "to_logfile", &logging_to_logfile, "off");
get_config_opt(config, local_handle, KEY_PREFIX "logfile", &logging_logfile,
"/var/log/pacemaker");
}
const_value = daemon_option("logfacility");
if (const_value) {
logging_syslog_facility = strdup(const_value);
crm_trace("Using logfacility setting from the environment: %s", logging_syslog_facility);
if (safe_str_eq(logging_syslog_facility, "none")) {
logging_to_syslog = strdup("off");
} else {
logging_to_syslog = strdup("on");
}
} else {
get_config_opt(config, local_handle, KEY_PREFIX "to_syslog", &logging_to_syslog, "on");
get_config_opt(config, local_handle, KEY_PREFIX "syslog_facility", &logging_syslog_facility,
"daemon");
}
#if HAVE_CONFDB
confdb_finalize(config);
#elif HAVE_CMAP
if(local_handle){
gid_t gid = 0;
if (crm_user_lookup(CRM_DAEMON_USER, NULL, &gid) < 0) {
crm_warn("No group found for user %s", CRM_DAEMON_USER);
} else {
char key[PATH_MAX];
snprintf(key, PATH_MAX, "uidgid.gid.%u", gid);
rc = cmap_set_uint8(local_handle, key, 1);
crm_notice("Configured corosync to accept connections from group %u: %s (%d)",
gid, ais_error2text(rc), rc);
}
}
cmap_finalize(local_handle);
#endif
if (daemon_option("debug")) {
crm_trace("Using debug setting from the environment: %s", daemon_option("debug"));
if (get_crm_log_level() < LOG_DEBUG && daemon_option_enabled("pacemakerd", "debug")) {
set_crm_log_level(LOG_DEBUG);
}
} else if (crm_is_true(logging_debug)) {
set_daemon_option("debug", "1");
if (get_crm_log_level() < LOG_DEBUG) {
set_crm_log_level(LOG_DEBUG);
}
} else {
set_daemon_option("debug", "0");
}
if (crm_is_true(logging_to_logfile)) {
if (crm_add_logfile(logging_logfile)) {
/* What a cluster fsck, eventually we need to mandate /one/ */
set_daemon_option("debugfile", logging_logfile);
set_daemon_option("DEBUGLOG", logging_logfile);
have_log = TRUE;
} else {
crm_err("Couldn't create logfile: %s", logging_logfile);
}
}
if (have_log && crm_is_true(logging_to_syslog) == FALSE) {
qb_log_ctl(QB_LOG_SYSLOG, QB_LOG_CONF_ENABLED, QB_FALSE);
free(logging_syslog_facility);
logging_syslog_facility = strdup("none");
crm_info("User configured file based logging and explicitly disabled syslog.");
} else if (crm_is_true(logging_to_syslog) == FALSE) {
crm_err("Please enable some sort of logging, either 'to_logfile: on' or 'to_syslog: on'.");
crm_err("If you use file logging, be sure to also define a value for 'logfile'");
}
set_daemon_option("logfacility", logging_syslog_facility);
setenv("HA_LOGFACILITY", logging_syslog_facility, 1);
free(logging_debug);
free(logging_logfile);
free(logging_to_logfile);
free(logging_to_syslog);
free(logging_syslog_facility);
return TRUE;
}
/* A_TE_INVOKE, A_TE_CANCEL */
void
do_te_invoke(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
if (AM_I_DC == FALSE || (fsa_state != S_TRANSITION_ENGINE && (action & A_TE_INVOKE))) {
crm_notice("No need to invoke the TE (%s) in state %s",
fsa_action2string(action), fsa_state2string(fsa_state));
return;
}
if (action & A_TE_CANCEL) {
crm_debug("Cancelling the transition: %s",
transition_graph->complete ? "inactive" : "active");
abort_transition(INFINITY, tg_restart, "Peer Cancelled", NULL);
if (transition_graph->complete == FALSE) {
crmd_fsa_stall(NULL);
}
} else if (action & A_TE_HALT) {
crm_debug("Halting the transition: %s", transition_graph->complete ? "inactive" : "active");
abort_transition(INFINITY, tg_stop, "Peer Halt", NULL);
if (transition_graph->complete == FALSE) {
crmd_fsa_stall(NULL);
}
} else if (action & A_TE_INVOKE) {
const char *value = NULL;
xmlNode *graph_data = NULL;
ha_msg_input_t *input = fsa_typed_data(fsa_dt_ha_msg);
const char *ref = crm_element_value(input->msg, XML_ATTR_REFERENCE);
const char *graph_file = crm_element_value(input->msg, F_CRM_TGRAPH);
const char *graph_input = crm_element_value(input->msg, F_CRM_TGRAPH_INPUT);
if (graph_file == NULL && input->xml == NULL) {
crm_log_xml_err(input->msg, "Bad command");
register_fsa_error(C_FSA_INTERNAL, I_FAIL, NULL);
return;
}
if (transition_graph->complete == FALSE) {
crm_info("Another transition is already active");
abort_transition(INFINITY, tg_restart, "Transition Active", NULL);
return;
}
if (fsa_pe_ref == NULL || safe_str_neq(fsa_pe_ref, ref)) {
crm_info("Transition is redundant: %s vs. %s", crm_str(fsa_pe_ref), crm_str(ref));
abort_transition(INFINITY, tg_restart, "Transition Redundant", NULL);
}
graph_data = input->xml;
if (graph_data == NULL && graph_file != NULL) {
graph_data = filename2xml(graph_file);
}
if (is_timer_started(transition_timer)) {
crm_debug("The transitioner wait for a transition timer");
return;
}
CRM_CHECK(graph_data != NULL,
crm_err("Input raised by %s is invalid", msg_data->origin);
crm_log_xml_err(input->msg, "Bad command");
return);
destroy_graph(transition_graph);
transition_graph = unpack_graph(graph_data, graph_input);
CRM_CHECK(transition_graph != NULL, transition_graph = create_blank_graph(); return);
crm_info("Processing graph %d (ref=%s) derived from %s", transition_graph->id, ref,
graph_input);
value = crm_element_value(graph_data, "failed-stop-offset");
if (value) {
free(failed_stop_offset);
failed_stop_offset = strdup(value);
}
value = crm_element_value(graph_data, "failed-start-offset");
if (value) {
free(failed_start_offset);
failed_start_offset = strdup(value);
}
trigger_graph();
print_graph(LOG_DEBUG_2, transition_graph);
if (graph_data != input->xml) {
free_xml(graph_data);
}
}
}
void
te_update_diff(const char *event, HA_Message *msg)
{
int rc = -1;
const char *op = NULL;
crm_data_t *diff = NULL;
crm_data_t *aborted = NULL;
const char *set_name = NULL;
int diff_add_updates = 0;
int diff_add_epoch = 0;
int diff_add_admin_epoch = 0;
int diff_del_updates = 0;
int diff_del_epoch = 0;
int diff_del_admin_epoch = 0;
if(msg == NULL) {
crm_err("NULL update");
return;
}
ha_msg_value_int(msg, F_CIB_RC, &rc);
op = cl_get_string(msg, F_CIB_OPERATION);
if(rc < cib_ok) {
crm_debug_2("Ignoring failed %s operation: %s",
op, cib_error2string(rc));
return;
}
diff = get_message_xml(msg, F_CIB_UPDATE_RESULT);
cib_diff_version_details(
diff,
&diff_add_admin_epoch, &diff_add_epoch, &diff_add_updates,
&diff_del_admin_epoch, &diff_del_epoch, &diff_del_updates);
crm_debug("Processing diff (%s): %d.%d.%d -> %d.%d.%d", op,
diff_del_admin_epoch,diff_del_epoch,diff_del_updates,
diff_add_admin_epoch,diff_add_epoch,diff_add_updates);
log_cib_diff(LOG_DEBUG_2, diff, op);
set_name = "diff-added";
if(diff != NULL) {
crm_data_t *section = NULL;
crm_data_t *change_set = find_xml_node(diff, set_name, FALSE);
change_set = find_xml_node(change_set, XML_TAG_CIB, FALSE);
if(change_set != NULL) {
crm_debug_2("Checking status changes");
section=get_object_root(XML_CIB_TAG_STATUS,change_set);
}
if(section != NULL) {
extract_event(section);
}
crm_debug_2("Checking change set: %s", set_name);
aborted = need_abort(change_set);
}
set_name = "diff-removed";
if(diff != NULL && aborted == NULL) {
crm_data_t *attrs = NULL;
crm_data_t *status = NULL;
crm_data_t *change_set = find_xml_node(diff, set_name, FALSE);
change_set = find_xml_node(change_set, XML_TAG_CIB, FALSE);
crm_debug_2("Checking change set: %s", set_name);
aborted = need_abort(change_set);
if(aborted == NULL && change_set != NULL) {
status = get_object_root(XML_CIB_TAG_STATUS, change_set);
xml_child_iter_filter(
status, node_state, XML_CIB_TAG_STATE,
attrs = find_xml_node(
node_state, XML_TAG_TRANSIENT_NODEATTRS, FALSE);
if(attrs != NULL) {
crm_info("Aborting on "XML_TAG_TRANSIENT_NODEATTRS" deletions");
abort_transition(INFINITY, tg_restart,
XML_TAG_TRANSIENT_NODEATTRS, attrs);
}
);
/* A_ELECTION_VOTE */
void
do_election_vote(long long action,
enum crmd_fsa_cause cause,
enum crmd_fsa_state cur_state,
enum crmd_fsa_input current_input, fsa_data_t * msg_data)
{
struct timeval age;
xmlNode *vote = NULL;
gboolean not_voting = FALSE;
/* don't vote if we're in one of these states or wanting to shut down */
switch (cur_state) {
case S_STARTING:
case S_RECOVERY:
case S_STOPPING:
case S_TERMINATE:
crm_warn("Not voting in election, we're in state %s", fsa_state2string(cur_state));
not_voting = TRUE;
break;
default:
break;
}
if (not_voting == FALSE) {
if (is_set(fsa_input_register, R_STARTING)) {
not_voting = TRUE;
}
}
if (not_voting) {
if (AM_I_DC) {
register_fsa_input(C_FSA_INTERNAL, I_RELEASE_DC, NULL);
} else {
register_fsa_input(C_FSA_INTERNAL, I_PENDING, NULL);
}
return;
}
vote = create_request(CRM_OP_VOTE, NULL, NULL, CRM_SYSTEM_CRMD, CRM_SYSTEM_CRMD, NULL);
current_election_id++;
crm_xml_add(vote, F_CRM_ELECTION_OWNER, fsa_our_uuid);
crm_xml_add_int(vote, F_CRM_ELECTION_ID, current_election_id);
crm_uptime(&age);
crm_xml_add_int(vote, F_CRM_ELECTION_AGE_S, age.tv_sec);
crm_xml_add_int(vote, F_CRM_ELECTION_AGE_US, age.tv_usec);
send_cluster_message(NULL, crm_msg_crmd, vote, TRUE);
free_xml(vote);
crm_debug("Started election %d", current_election_id);
if (voted) {
g_hash_table_destroy(voted);
}
voted = NULL;
if (cur_state == S_ELECTION || cur_state == S_RELEASE_DC) {
crm_timer_start(election_timeout);
} else if (cur_state != S_INTEGRATION) {
crm_err("Broken? Voting in state %s", fsa_state2string(cur_state));
}
return;
}
int
main(int argc, char **argv)
{
int option_index = 0;
int argerr = 0;
int flag;
crm_log_init(NULL, LOG_ERR, FALSE, TRUE, argc, argv);
crm_set_options(NULL, "command [options]", long_options,
"Development tool for performing some crmd-specific commands."
"\n Likely to be replaced by crm_node in the future");
if (argc < 2) {
crm_help('?', LSB_EXIT_EINVAL);
}
while (1) {
flag = crm_get_option(argc, argv, &option_index);
if (flag == -1)
break;
switch (flag) {
case 'V':
BE_VERBOSE = TRUE;
admin_verbose = XML_BOOLEAN_TRUE;
crm_bump_log_level();
break;
case 't':
message_timeout_ms = atoi(optarg);
if (message_timeout_ms < 1) {
message_timeout_ms = 30 * 1000;
}
break;
case '$':
case '?':
crm_help(flag, LSB_EXIT_OK);
break;
case 'D':
DO_WHOIS_DC = TRUE;
break;
case 'B':
BASH_EXPORT = TRUE;
break;
case 'K':
DO_RESET = TRUE;
crm_debug_2("Option %c => %s", flag, optarg);
dest_node = crm_strdup(optarg);
crmd_operation = CRM_OP_LOCAL_SHUTDOWN;
break;
case 'q':
BE_SILENT = TRUE;
break;
case 'i':
DO_DEBUG = debug_inc;
crm_debug_2("Option %c => %s", flag, optarg);
dest_node = crm_strdup(optarg);
break;
case 'd':
DO_DEBUG = debug_dec;
crm_debug_2("Option %c => %s", flag, optarg);
dest_node = crm_strdup(optarg);
break;
case 'S':
DO_HEALTH = TRUE;
crm_debug_2("Option %c => %s", flag, optarg);
dest_node = crm_strdup(optarg);
break;
case 'E':
DO_ELECT_DC = TRUE;
break;
case 'N':
DO_NODE_LIST = TRUE;
break;
case 'H':
DO_HEALTH = TRUE;
break;
default:
printf("Argument code 0%o (%c) is not (?yet?) supported\n", flag, flag);
++argerr;
break;
}
}
if (optind < argc) {
printf("non-option ARGV-elements: ");
while (optind < argc)
printf("%s ", argv[optind++]);
printf("\n");
}
if (optind > argc) {
++argerr;
}
if (argerr) {
crm_help('?', LSB_EXIT_GENERIC);
}
if (do_init()) {
int res = 0;
res = do_work();
if (res > 0) {
/* wait for the reply by creating a mainloop and running it until
* the callbacks are invoked...
*/
mainloop = g_main_new(FALSE);
expected_responses++;
crm_debug_2("Waiting for %d replies from the local CRM", expected_responses);
message_timer_id = g_timeout_add(message_timeout_ms, admin_message_timeout, NULL);
g_main_run(mainloop);
} else if (res < 0) {
crm_err("No message to send");
operation_status = -1;
}
} else {
crm_warn("Init failed, could not perform requested operations");
operation_status = -2;
}
crm_debug_2("%s exiting normally", crm_system_name);
return operation_status;
}
static void
stonith_peer_cs_destroy(gpointer user_data)
{
crm_err("Corosync connection terminated");
stonith_shutdown(0);
}
int
do_work(void)
{
int ret = 1;
/* construct the request */
xmlNode *msg_data = NULL;
gboolean all_is_good = TRUE;
msg_options = create_xml_node(NULL, XML_TAG_OPTIONS);
crm_xml_add(msg_options, XML_ATTR_VERBOSE, admin_verbose);
crm_xml_add(msg_options, XML_ATTR_TIMEOUT, "0");
if (DO_HEALTH == TRUE) {
crm_debug_2("Querying the system");
sys_to = CRM_SYSTEM_DC;
if (dest_node != NULL) {
sys_to = CRM_SYSTEM_CRMD;
crmd_operation = CRM_OP_PING;
if (BE_VERBOSE) {
expected_responses = 1;
}
crm_xml_add(msg_options, XML_ATTR_TIMEOUT, "0");
} else {
crm_info("Cluster-wide health not available yet");
all_is_good = FALSE;
}
} else if (DO_ELECT_DC) {
/* tell the local node to initiate an election */
sys_to = CRM_SYSTEM_CRMD;
crmd_operation = CRM_OP_VOTE;
crm_xml_add(msg_options, XML_ATTR_TIMEOUT, "0");
dest_node = NULL;
ret = 0; /* no return message */
} else if (DO_WHOIS_DC) {
sys_to = CRM_SYSTEM_DC;
crmd_operation = CRM_OP_PING;
crm_xml_add(msg_options, XML_ATTR_TIMEOUT, "0");
dest_node = NULL;
} else if (DO_NODE_LIST) {
cib_t *the_cib = cib_new();
xmlNode *output = NULL;
enum cib_errors rc = the_cib->cmds->signon(the_cib, crm_system_name, cib_command);
if (rc != cib_ok) {
return -1;
}
output = get_cib_copy(the_cib);
do_find_node_list(output);
free_xml(output);
the_cib->cmds->signoff(the_cib);
exit(rc);
} else if (DO_RESET) {
/* tell dest_node to initiate the shutdown proceedure
*
* if dest_node is NULL, the request will be sent to the
* local node
*/
sys_to = CRM_SYSTEM_CRMD;
crm_xml_add(msg_options, XML_ATTR_TIMEOUT, "0");
ret = 0; /* no return message */
} else if (DO_DEBUG == debug_inc) {
/* tell dest_node to increase its debug level
*
* if dest_node is NULL, the request will be sent to the
* local node
*/
sys_to = CRM_SYSTEM_CRMD;
crmd_operation = CRM_OP_DEBUG_UP;
ret = 0; /* no return message */
} else if (DO_DEBUG == debug_dec) {
/* tell dest_node to increase its debug level
*
* if dest_node is NULL, the request will be sent to the
* local node
*/
sys_to = CRM_SYSTEM_CRMD;
crmd_operation = CRM_OP_DEBUG_DOWN;
ret = 0; /* no return message */
} else {
crm_err("Unknown options");
all_is_good = FALSE;
}
if (all_is_good == FALSE) {
crm_err("Creation of request failed. No message to send");
return -1;
}
/* send it */
if (crmd_channel == NULL) {
crm_err("The IPC connection is not valid, cannot send anything");
return -1;
}
if (sys_to == NULL) {
if (dest_node != NULL) {
sys_to = CRM_SYSTEM_CRMD;
} else {
sys_to = CRM_SYSTEM_DC;
}
}
{
xmlNode *cmd = create_request(crmd_operation, msg_data, dest_node, sys_to,
crm_system_name, admin_uuid);
send_ipc_message(crmd_channel, cmd);
free_xml(cmd);
}
return ret;
}
svc_action_t *resources_action_create(
const char *name, const char *standard, const char *provider, const char *agent,
const char *action, int interval, int timeout, GHashTable *params)
{
svc_action_t *op;
/*
* Do some up front sanity checks before we go off and
* build the svc_action_t instance.
*/
if (crm_strlen_zero(name)) {
crm_err("A service or resource action must have a name.");
return NULL;
}
if (crm_strlen_zero(standard)) {
crm_err("A service action must have a valid standard.");
return NULL;
}
if (!strcasecmp(standard, "ocf") && crm_strlen_zero(provider)) {
crm_err("An OCF resource action must have a provider.");
return NULL;
}
if (crm_strlen_zero(agent)) {
crm_err("A service or resource action must have an agent.");
return NULL;
}
if (crm_strlen_zero(action)) {
crm_err("A service or resource action must specify an action.");
return NULL;
}
if (safe_str_eq(action, "monitor") && (safe_str_eq(standard, "lsb") || safe_str_eq(standard, "service"))) {
action = "status";
}
/*
* Sanity checks passed, proceed!
*/
op = calloc(1, sizeof(svc_action_t));
op->opaque = calloc(1, sizeof(svc_action_private_t));
op->rsc = strdup(name);
op->action = strdup(action);
op->interval = interval;
op->timeout = timeout;
op->standard = strdup(standard);
op->agent = strdup(agent);
op->sequence = ++operations;
if (asprintf(&op->id, "%s_%s_%d", name, action, interval) == -1) {
goto return_error;
}
if(strcasecmp(op->standard, "service") == 0) {
/* Work it out and then fall into the if-else block below.
* Priority is:
* - lsb
* - systemd
* - upstart
*/
int rc = 0;
struct stat st;
char *path = NULL;
#ifdef LSB_ROOT_DIR
rc = asprintf(&path, "%s/%s", LSB_ROOT_DIR, op->agent);
if(rc > 0 && stat(path, &st) == 0) {
crm_debug("Found an lsb agent for %s/% the", op->rsc, op->agent);
free(path);
free(op->standard);
op->standard = strdup("lsb");
goto expanded;
}
free(path);
#endif
#if SUPPORT_SYSTEMD
if(systemd_unit_exists(op->agent)) {
crm_debug("Found a systemd agent for %s/%s", op->rsc, op->agent);
free(op->standard);
op->standard = strdup("systemd");
goto expanded;
}
#endif
#if SUPPORT_UPSTART
if(upstart_job_exists(op->agent)) {
crm_debug("Found an upstart agent for %s/%s", op->rsc, op->agent);
free(op->standard);
op->standard = strdup("upstart");
goto expanded;
}
#endif
crm_info("Cannot determine the standard for %s (%s)", op->rsc, op->agent);
}
expanded:
if(strcasecmp(op->standard, "ocf") == 0) {
op->provider = strdup(provider);
op->params = params;
if (asprintf(&op->opaque->exec, "%s/resource.d/%s/%s",
OCF_ROOT_DIR, provider, agent) == -1) {
goto return_error;
}
op->opaque->args[0] = strdup(op->opaque->exec);
op->opaque->args[1] = strdup(action);
} else if(strcasecmp(op->standard, "lsb") == 0) {
if (op->agent[0] == '/') {
/* if given an absolute path, use that instead
* of tacking on the LSB_ROOT_DIR path to the front */
op->opaque->exec = strdup(op->agent);
} else if (asprintf(&op->opaque->exec, "%s/%s", LSB_ROOT_DIR, op->agent) == -1) {
goto return_error;
}
op->opaque->args[0] = strdup(op->opaque->exec);
op->opaque->args[1] = strdup(op->action);
op->opaque->args[2] = NULL;
#if SUPPORT_SYSTEMD
} else if(strcasecmp(op->standard, "systemd") == 0) {
op->opaque->exec = strdup("systemd-dbus");
#endif
#if SUPPORT_UPSTART
} else if(strcasecmp(op->standard, "upstart") == 0) {
op->opaque->exec = strdup("upstart-dbus");
#endif
} else if(strcasecmp(op->standard, "service") == 0) {
op->opaque->exec = strdup(SERVICE_SCRIPT);
op->opaque->args[0] = strdup(SERVICE_SCRIPT);
op->opaque->args[1] = strdup(agent);
op->opaque->args[2] = strdup(action);
} else {
crm_err("Unknown resource standard: %s", op->standard);
services_action_free(op);
op = NULL;
}
return op;
return_error:
services_action_free(op);
return NULL;
}
gboolean
relay_message(xmlNode * msg, gboolean originated_locally)
{
int dest = 1;
int is_for_dc = 0;
int is_for_dcib = 0;
int is_for_te = 0;
int is_for_crm = 0;
int is_for_cib = 0;
int is_local = 0;
gboolean processing_complete = FALSE;
const char *host_to = crm_element_value(msg, F_CRM_HOST_TO);
const char *sys_to = crm_element_value(msg, F_CRM_SYS_TO);
const char *sys_from = crm_element_value(msg, F_CRM_SYS_FROM);
const char *type = crm_element_value(msg, F_TYPE);
const char *msg_error = NULL;
crm_trace("Routing message %s", crm_element_value(msg, XML_ATTR_REFERENCE));
if (msg == NULL) {
msg_error = "Cannot route empty message";
} else if (safe_str_eq(CRM_OP_HELLO, crm_element_value(msg, F_CRM_TASK))) {
/* quietly ignore */
processing_complete = TRUE;
} else if (safe_str_neq(type, T_CRM)) {
msg_error = "Bad message type";
} else if (sys_to == NULL) {
msg_error = "Bad message destination: no subsystem";
}
if (msg_error != NULL) {
processing_complete = TRUE;
crm_err("%s", msg_error);
crm_log_xml_warn(msg, "bad msg");
}
if (processing_complete) {
return TRUE;
}
processing_complete = TRUE;
is_for_dc = (strcasecmp(CRM_SYSTEM_DC, sys_to) == 0);
is_for_dcib = (strcasecmp(CRM_SYSTEM_DCIB, sys_to) == 0);
is_for_te = (strcasecmp(CRM_SYSTEM_TENGINE, sys_to) == 0);
is_for_cib = (strcasecmp(CRM_SYSTEM_CIB, sys_to) == 0);
is_for_crm = (strcasecmp(CRM_SYSTEM_CRMD, sys_to) == 0);
is_local = 0;
if (host_to == NULL || strlen(host_to) == 0) {
if (is_for_dc || is_for_te) {
is_local = 0;
} else if (is_for_crm && originated_locally) {
is_local = 0;
} else {
is_local = 1;
}
} else if (safe_str_eq(fsa_our_uname, host_to)) {
is_local = 1;
}
if (is_for_dc || is_for_dcib || is_for_te) {
if (AM_I_DC && is_for_te) {
ROUTER_RESULT("Message result: Local relay");
send_msg_via_ipc(msg, sys_to);
} else if (AM_I_DC) {
ROUTER_RESULT("Message result: DC/CRMd process");
processing_complete = FALSE; /* more to be done by caller */
} else if (originated_locally && safe_str_neq(sys_from, CRM_SYSTEM_PENGINE)
&& safe_str_neq(sys_from, CRM_SYSTEM_TENGINE)) {
/* Neither the TE or PE should be sending messages
* to DC's on other nodes
*
* By definition, if we are no longer the DC, then
* the PE or TE's data should be discarded
*/
#if SUPPORT_COROSYNC
if (is_openais_cluster()) {
dest = text2msg_type(sys_to);
}
#endif
ROUTER_RESULT("Message result: External relay to DC");
send_cluster_message(host_to, dest, msg, TRUE);
} else {
/* discard */
ROUTER_RESULT("Message result: Discard, not DC");
}
} else if (is_local && (is_for_crm || is_for_cib)) {
ROUTER_RESULT("Message result: CRMd process");
processing_complete = FALSE; /* more to be done by caller */
} else if (is_local) {
ROUTER_RESULT("Message result: Local relay");
send_msg_via_ipc(msg, sys_to);
} else {
#if SUPPORT_COROSYNC
if (is_openais_cluster()) {
dest = text2msg_type(sys_to);
}
#endif
ROUTER_RESULT("Message result: External relay");
send_cluster_message(host_to, dest, msg, TRUE);
}
return processing_complete;
}