qd_connector_t *qd_dispatch_configure_connector(qd_dispatch_t *qd, qd_entity_t *entity)
{
qd_connection_manager_t *cm = qd->connection_manager;
qd_connector_t *ct = qd_server_connector(qd->server);
if (ct && load_server_config(qd, &ct->config, entity, false) == QD_ERROR_NONE) {
DEQ_ITEM_INIT(ct);
DEQ_INSERT_TAIL(cm->connectors, ct);
log_config(cm->log_source, &ct->config, "Connector");
//
// Add the first item to the ct->conn_info_list
// The initial connection information and any backup connection information is stored in the conn_info_list
//
qd_failover_item_t *item = NEW(qd_failover_item_t);
ZERO(item);
item->scheme = 0;
item->host = strdup(ct->config.host);
item->port = strdup(ct->config.port);
item->hostname = 0;
int hplen = strlen(item->host) + strlen(item->port) + 2;
item->host_port = malloc(hplen);
snprintf(item->host_port, hplen, "%s:%s", item->host , item->port);
DEQ_INSERT_TAIL(ct->conn_info_list, item);
return ct;
}
qd_log(cm->log_source, QD_LOG_ERROR, "Unable to create connector: %s", qd_error_message());
qd_connector_decref(ct);
return 0;
}
void start_core(int argc, char **argv) {
int orig_argc = argc;
char **orig_argv = copy_args(argc, argv);
FLAGS_stderrthreshold = 0;
FLAGS_logbuflevel = -1;
google::ParseCommandLineFlags(&argc, &argv, true);
google::InitGoogleLogging(argv[0]);
auto conf = create_config();
ld_environment(orig_argv, conf.rules_directory);
free_args(orig_argc, orig_argv);
log_config(conf);
g_core = make_real_core(conf);
g_core->start();
g_core.reset();
free_thread_ns();
}
qd_listener_t *qd_dispatch_configure_listener(qd_dispatch_t *qd, qd_entity_t *entity)
{
qd_connection_manager_t *cm = qd->connection_manager;
qd_listener_t *li = qd_server_listener(qd->server);
if (!li || load_server_config(qd, &li->config, entity, true) != QD_ERROR_NONE) {
qd_log(cm->log_source, QD_LOG_ERROR, "Unable to create listener: %s", qd_error_message());
qd_listener_decref(li);
return 0;
}
char *fol = qd_entity_opt_string(entity, "failoverList", 0);
if (fol) {
li->config.failover_list = qd_failover_list(fol);
free(fol);
if (li->config.failover_list == 0) {
qd_log(cm->log_source, QD_LOG_ERROR, "Unable to create listener, bad failover list: %s",
qd_error_message());
qd_listener_decref(li);
return 0;
}
} else {
li->config.failover_list = 0;
}
DEQ_ITEM_INIT(li);
DEQ_INSERT_TAIL(cm->listeners, li);
log_config(cm->log_source, &li->config, "Listener");
return li;
}
/* main - slurmctld main function, start various threads and process RPCs */
int main(int argc, char *argv[])
{
pthread_attr_t thread_attr;
char node_name[128];
void *db_conn = NULL;
assoc_init_args_t assoc_init_arg;
_init_config();
log_init(argv[0], log_opts, LOG_DAEMON, NULL);
if (read_slurmdbd_conf())
exit(1);
_parse_commandline(argc, argv);
_update_logging(true);
_update_nice();
if (slurm_auth_init(NULL) != SLURM_SUCCESS) {
fatal("Unable to initialize %s authentication plugin",
slurmdbd_conf->auth_type);
}
if (slurm_acct_storage_init(NULL) != SLURM_SUCCESS) {
fatal("Unable to initialize %s accounting storage plugin",
slurmdbd_conf->storage_type);
}
_kill_old_slurmdbd();
if (foreground == 0)
_daemonize();
/*
* Need to create pidfile here in case we setuid() below
* (init_pidfile() exits if it can't initialize pid file).
* On Linux we also need to make this setuid job explicitly
* able to write a core dump.
* This also has to happen after daemon(), which closes all fd's,
* so we keep the write lock of the pidfile.
*/
_init_pidfile();
_become_slurm_user();
if (foreground == 0)
_set_work_dir();
log_config();
#ifdef PR_SET_DUMPABLE
if (prctl(PR_SET_DUMPABLE, 1) < 0)
debug ("Unable to set dumpable to 1");
#endif /* PR_SET_DUMPABLE */
if (xsignal_block(dbd_sigarray) < 0)
error("Unable to block signals");
/* Create attached thread for signal handling */
slurm_attr_init(&thread_attr);
if (pthread_create(&signal_handler_thread, &thread_attr,
_signal_handler, NULL))
fatal("pthread_create %m");
slurm_attr_destroy(&thread_attr);
registered_clusters = list_create(NULL);
slurm_attr_init(&thread_attr);
if (pthread_create(&commit_handler_thread, &thread_attr,
_commit_handler, NULL))
fatal("pthread_create %m");
slurm_attr_destroy(&thread_attr);
memset(&assoc_init_arg, 0, sizeof(assoc_init_args_t));
/* If we are tacking wckey we need to cache
wckeys, if we aren't only cache the users, qos */
assoc_init_arg.cache_level = ASSOC_MGR_CACHE_USER | ASSOC_MGR_CACHE_QOS;
if (slurmdbd_conf->track_wckey)
assoc_init_arg.cache_level |= ASSOC_MGR_CACHE_WCKEY;
db_conn = acct_storage_g_get_connection(NULL, 0, true, NULL);
if (assoc_mgr_init(db_conn, &assoc_init_arg, errno) == SLURM_ERROR) {
error("Problem getting cache of data");
acct_storage_g_close_connection(&db_conn);
goto end_it;
}
if (gethostname_short(node_name, sizeof(node_name)))
fatal("getnodename: %m");
while (1) {
if (slurmdbd_conf->dbd_backup &&
(!strcmp(node_name, slurmdbd_conf->dbd_backup) ||
!strcmp(slurmdbd_conf->dbd_backup, "localhost"))) {
info("slurmdbd running in background mode");
have_control = false;
backup = true;
/* make sure any locks are released */
acct_storage_g_commit(db_conn, 1);
run_dbd_backup();
if (!shutdown_time)
assoc_mgr_refresh_lists(db_conn);
} else if (slurmdbd_conf->dbd_host &&
(!strcmp(slurmdbd_conf->dbd_host, node_name) ||
!strcmp(slurmdbd_conf->dbd_host, "localhost"))) {
backup = false;
have_control = true;
} else {
fatal("This host not configured to run SlurmDBD "
"(%s != %s | (backup) %s)",
node_name, slurmdbd_conf->dbd_host,
slurmdbd_conf->dbd_backup);
}
if (!shutdown_time) {
/* Create attached thread to process incoming RPCs */
slurm_attr_init(&thread_attr);
if (pthread_create(&rpc_handler_thread, &thread_attr,
rpc_mgr, NULL))
fatal("pthread_create error %m");
slurm_attr_destroy(&thread_attr);
}
if (!shutdown_time) {
/* Create attached thread to do usage rollup */
slurm_attr_init(&thread_attr);
if (pthread_create(&rollup_handler_thread,
&thread_attr,
_rollup_handler, db_conn))
fatal("pthread_create error %m");
slurm_attr_destroy(&thread_attr);
}
/* Daemon is fully operational here */
if (!shutdown_time || primary_resumed) {
shutdown_time = 0;
info("slurmdbd version %s started",
SLURM_VERSION_STRING);
if (backup)
run_dbd_backup();
}
_request_registrations(db_conn);
acct_storage_g_commit(db_conn, 1);
/* this is only ran if not backup */
if (rollup_handler_thread)
pthread_join(rollup_handler_thread, NULL);
if (rpc_handler_thread)
pthread_join(rpc_handler_thread, NULL);
if (backup && primary_resumed) {
shutdown_time = 0;
info("Backup has given up control");
}
if (shutdown_time)
break;
}
/* Daemon termination handled here */
end_it:
if (signal_handler_thread)
pthread_join(signal_handler_thread, NULL);
if (commit_handler_thread)
pthread_join(commit_handler_thread, NULL);
acct_storage_g_commit(db_conn, 1);
acct_storage_g_close_connection(&db_conn);
if (slurmdbd_conf->pid_file &&
(unlink(slurmdbd_conf->pid_file) < 0)) {
verbose("Unable to remove pidfile '%s': %m",
slurmdbd_conf->pid_file);
}
FREE_NULL_LIST(registered_clusters);
assoc_mgr_fini(NULL);
slurm_acct_storage_fini();
slurm_auth_fini();
log_fini();
free_slurmdbd_conf();
exit(0);
}