int
orte_sds_slurm_set_name(void)
{
int rc;
int id;
int vpid_start;
int num_procs;
char* name_string = NULL;
int slurm_nodeid;
/* start by getting our cellid, jobid, and vpid (which is the
starting vpid for the list of daemons) */
id = mca_base_param_register_string("ns", "nds", "name", NULL, NULL);
mca_base_param_lookup_string(id, &name_string);
if(name_string != NULL) {
if (ORTE_SUCCESS !=
(rc = orte_ns.convert_string_to_process_name(&(orte_process_info.my_name),
name_string))) {
ORTE_ERROR_LOG(rc);
free(name_string);
return rc;
}
free(name_string);
} else {
orte_cellid_t cellid;
orte_jobid_t jobid;
orte_vpid_t vpid;
char* cellid_string;
char* jobid_string;
char* vpid_string;
id = mca_base_param_register_string("ns", "nds", "cellid", NULL, NULL);
mca_base_param_lookup_string(id, &cellid_string);
if (NULL == cellid_string) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_ns.convert_string_to_cellid(&cellid, cellid_string))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
id = mca_base_param_register_string("ns", "nds", "jobid", NULL, NULL);
mca_base_param_lookup_string(id, &jobid_string);
if (NULL == jobid_string) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_ns.convert_string_to_jobid(&jobid, jobid_string))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
id = mca_base_param_register_string("ns", "nds", "vpid", NULL, NULL);
mca_base_param_lookup_string(id, &vpid_string);
if (NULL == vpid_string) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_ns.convert_string_to_vpid(&vpid, vpid_string))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
if (ORTE_SUCCESS != (rc = orte_ns.create_process_name(&(orte_process_info.my_name),
cellid,
jobid,
vpid))) {
ORTE_ERROR_LOG(rc);
return rc;
}
}
/* fix up the base name and make it the "real" name */
slurm_nodeid = atoi(getenv("SLURM_NODEID"));
orte_process_info.my_name->vpid += slurm_nodeid;
/* fix up the system info nodename to match exactly what slurm returned */
if (NULL != orte_system_info.nodename) {
free(orte_system_info.nodename);
}
orte_system_info.nodename = get_slurm_nodename(slurm_nodeid);
id = mca_base_param_register_int("ns", "nds", "vpid_start", NULL, -1);
mca_base_param_lookup_int(id, &vpid_start);
if (vpid_start < 0) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
id = mca_base_param_register_int("ns", "nds", "num_procs", NULL, -1);
mca_base_param_lookup_int(id, &num_procs);
if (num_procs < 0) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
orte_process_info.vpid_start = (orte_vpid_t)vpid_start;
orte_process_info.num_procs = (size_t)num_procs;
return ORTE_SUCCESS;
}
static int slurm_set_name(void)
{
int slurm_nodeid;
int rc;
orte_jobid_t jobid;
orte_vpid_t vpid;
char* tmp;
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:slurm setting name"));
mca_base_param_reg_string_name("orte", "ess_jobid", "Process jobid",
true, false, NULL, &tmp);
if (NULL == tmp) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_jobid(&jobid, tmp))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
free(tmp);
mca_base_param_reg_string_name("orte", "ess_vpid", "Process vpid",
true, false, NULL, &tmp);
if (NULL == tmp) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_vpid(&vpid, tmp))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
free(tmp);
ORTE_PROC_MY_NAME->jobid = jobid;
/* fix up the vpid and make it the "real" vpid */
slurm_nodeid = atoi(getenv("SLURM_NODEID"));
ORTE_PROC_MY_NAME->vpid = vpid + slurm_nodeid;
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:slurm set name to %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* get my node rank in case we are using static ports - this won't
* be present for daemons, so don't error out if we don't have it
*/
mca_base_param_reg_string_name("orte", "ess_node_rank", "Process node rank",
true, false, NULL, &tmp);
if (NULL != tmp) {
my_node_rank = strtol(tmp, NULL, 10);
}
/* fix up the system info nodename to match exactly what slurm returned */
if (NULL != orte_process_info.nodename) {
free(orte_process_info.nodename);
}
orte_process_info.nodename = get_slurm_nodename(slurm_nodeid);
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:slurm set nodename to %s",
orte_process_info.nodename));
/* get the non-name common environmental variables */
if (ORTE_SUCCESS != (rc = orte_ess_env_get())) {
ORTE_ERROR_LOG(rc);
return rc;
}
return ORTE_SUCCESS;
}
static int slurm_set_name(void)
{
int slurm_nodeid;
int rc;
int id;
orte_jobid_t jobid;
orte_vpid_t vpid;
char* jobid_string;
char* vpid_string;
char *nodeid;
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:slurm setting name"));
id = mca_base_param_register_string("orte", "ess", "jobid", NULL, NULL);
mca_base_param_lookup_string(id, &jobid_string);
if (NULL == jobid_string) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_jobid(&jobid, jobid_string))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
id = mca_base_param_register_string("orte", "ess", "vpid", NULL, NULL);
mca_base_param_lookup_string(id, &vpid_string);
if (NULL == vpid_string) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
return ORTE_ERR_NOT_FOUND;
}
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_vpid(&vpid, vpid_string))) {
ORTE_ERROR_LOG(rc);
return(rc);
}
ORTE_PROC_MY_NAME->jobid = jobid;
/* fix up the vpid and make it the "real" vpid */
if (NULL == (nodeid = getenv("SLURM_NODEID"))) {
opal_output(0, "SLURM_NODEID not found - cannot define name");
return ORTE_ERR_NOT_FOUND;
}
slurm_nodeid = atoi(nodeid);
ORTE_PROC_MY_NAME->vpid = vpid + slurm_nodeid;
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:slurm set name to %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* fix up the system info nodename to match exactly what slurm returned */
if (NULL != orte_process_info.nodename) {
free(orte_process_info.nodename);
}
orte_process_info.nodename = get_slurm_nodename(slurm_nodeid);
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_output,
"ess:slurm set nodename to %s",
orte_process_info.nodename));
/* get the non-name common environmental variables */
if (ORTE_SUCCESS != (rc = orte_ess_env_get())) {
ORTE_ERROR_LOG(rc);
return rc;
}
return ORTE_SUCCESS;
}
