/* Initialize power_save module parameters.
* Return 0 on valid configuration to run power saving,
* otherwise log the problem and return -1 */
static int _init_power_config(void)
{
slurm_ctl_conf_t *conf = slurm_conf_lock();
last_config = slurmctld_conf.last_update;
idle_time = conf->suspend_time - 1;
suspend_rate = conf->suspend_rate;
resume_timeout = conf->resume_timeout;
resume_rate = conf->resume_rate;
slurmd_timeout = conf->slurmd_timeout;
suspend_timeout = conf->suspend_timeout;
_clear_power_config();
if (conf->suspend_program)
suspend_prog = xstrdup(conf->suspend_program);
if (conf->resume_program)
resume_prog = xstrdup(conf->resume_program);
if (conf->suspend_exc_nodes)
exc_nodes = xstrdup(conf->suspend_exc_nodes);
if (conf->suspend_exc_parts)
exc_parts = xstrdup(conf->suspend_exc_parts);
slurm_conf_unlock();
if (idle_time < 0) { /* not an error */
debug("power_save module disabled, SuspendTime < 0");
return -1;
}
if (suspend_rate < 0) {
error("power_save module disabled, SuspendRate < 0");
return -1;
}
if (resume_rate < 0) {
error("power_save module disabled, ResumeRate < 0");
return -1;
}
if (suspend_prog == NULL) {
error("power_save module disabled, NULL SuspendProgram");
return -1;
} else if (!_valid_prog(suspend_prog)) {
error("power_save module disabled, invalid SuspendProgram %s",
suspend_prog);
return -1;
}
if (resume_prog == NULL) {
error("power_save module disabled, NULL ResumeProgram");
return -1;
} else if (!_valid_prog(resume_prog)) {
error("power_save module disabled, invalid ResumeProgram %s",
resume_prog);
return -1;
}
return 0;
}
/*
* init_power_save - Initialize the power save module. Started as a
* pthread. Terminates automatically at slurmctld shutdown time.
* Input and output are unused.
*/
static void *_init_power_save(void *arg)
{
/* Locks: Read nodes */
slurmctld_lock_t node_read_lock = {
NO_LOCK, READ_LOCK, NO_LOCK, NO_LOCK };
/* Locks: Write nodes */
slurmctld_lock_t node_write_lock = {
NO_LOCK, WRITE_LOCK, NO_LOCK, NO_LOCK };
time_t now, boot_time = 0, last_power_scan = 0;
if (power_save_config && !power_save_enabled) {
debug("power_save mode not enabled");
return NULL;
}
suspend_node_bitmap = bit_alloc(node_record_count);
resume_node_bitmap = bit_alloc(node_record_count);
while (slurmctld_config.shutdown_time == 0) {
sleep(1);
if (_reap_procs() < 2) {
debug("power_save programs getting backlogged");
continue;
}
if ((last_config != slurmctld_conf.last_update) &&
(_init_power_config())) {
info("power_save mode has been disabled due to "
"configuration changes");
goto fini;
}
now = time(NULL);
if (boot_time == 0)
boot_time = now;
/* Only run every 60 seconds or after a node state change,
* whichever happens first */
if ((last_node_update >= last_power_scan) ||
(now >= (last_power_scan + 60))) {
lock_slurmctld(node_write_lock);
_do_power_work(now);
unlock_slurmctld(node_write_lock);
last_power_scan = now;
}
if (slurmd_timeout &&
(now > (boot_time + (slurmd_timeout / 2)))) {
lock_slurmctld(node_read_lock);
_re_wake();
unlock_slurmctld(node_read_lock);
/* prevent additional executions */
boot_time += (365 * 24 * 60 * 60);
slurmd_timeout = 0;
}
}
fini: _clear_power_config();
FREE_NULL_BITMAP(suspend_node_bitmap);
FREE_NULL_BITMAP(resume_node_bitmap);
_shutdown_power();
slurm_mutex_lock(&power_mutex);
power_save_enabled = false;
pthread_cond_signal(&power_cond);
slurm_mutex_unlock(&power_mutex);
pthread_exit(NULL);
return NULL;
}
/* Initialize power_save module parameters.
* Return 0 on valid configuration to run power saving,
* otherwise log the problem and return -1 */
static int _init_power_config(void)
{
slurm_ctl_conf_t *conf = slurm_conf_lock();
last_config = slurmctld_conf.last_update;
idle_time = conf->suspend_time - 1;
suspend_rate = conf->suspend_rate;
resume_timeout = conf->resume_timeout;
resume_rate = conf->resume_rate;
slurmd_timeout = conf->slurmd_timeout;
suspend_timeout = conf->suspend_timeout;
_clear_power_config();
if (conf->suspend_program)
suspend_prog = xstrdup(conf->suspend_program);
if (conf->resume_program)
resume_prog = xstrdup(conf->resume_program);
if (conf->suspend_exc_nodes)
exc_nodes = xstrdup(conf->suspend_exc_nodes);
if (conf->suspend_exc_parts)
exc_parts = xstrdup(conf->suspend_exc_parts);
slurm_conf_unlock();
if (idle_time < 0) { /* not an error */
debug("power_save module disabled, SuspendTime < 0");
return -1;
}
if (suspend_rate < 0) {
error("power_save module disabled, SuspendRate < 0");
return -1;
}
if (resume_rate < 0) {
error("power_save module disabled, ResumeRate < 0");
return -1;
}
if (suspend_prog == NULL) {
error("power_save module disabled, NULL SuspendProgram");
return -1;
} else if (!_valid_prog(suspend_prog)) {
error("power_save module disabled, invalid SuspendProgram %s",
suspend_prog);
return -1;
}
if (resume_prog == NULL) {
error("power_save module disabled, NULL ResumeProgram");
return -1;
} else if (!_valid_prog(resume_prog)) {
error("power_save module disabled, invalid ResumeProgram %s",
resume_prog);
return -1;
}
if (exc_nodes &&
(node_name2bitmap(exc_nodes, false, &exc_node_bitmap))) {
error("power_save module disabled, "
"invalid SuspendExcNodes %s", exc_nodes);
return -1;
}
if (exc_parts) {
char *tmp = NULL, *one_part = NULL, *part_list = NULL;
struct part_record *part_ptr = NULL;
int rc = 0;
part_list = xstrdup(exc_parts);
one_part = strtok_r(part_list, ",", &tmp);
while (one_part != NULL) {
part_ptr = find_part_record(one_part);
if (!part_ptr) {
error("power_save module disabled, "
"invalid SuspendExcPart %s",
one_part);
rc = -1;
break;
}
if (exc_node_bitmap)
bit_or(exc_node_bitmap, part_ptr->node_bitmap);
else
exc_node_bitmap = bit_copy(part_ptr->
node_bitmap);
one_part = strtok_r(NULL, ",", &tmp);
}
xfree(part_list);
if (rc)
return rc;
}
if (exc_node_bitmap) {
char *tmp = bitmap2node_name(exc_node_bitmap);
debug("power_save module, excluded nodes %s", tmp);
xfree(tmp);
}
return 0;
}