static void _set_node_mixed(node_info_msg_t *resp)
{
node_info_t *node_ptr = NULL;
int i;
if (!resp)
return;
for (i = 0, node_ptr = resp->node_array;
i < resp->record_count; i++, node_ptr++) {
uint16_t used_cpus = 0;
select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ALLOCATED, &used_cpus);
if ((used_cpus != 0) && (used_cpus != node_ptr->cpus)) {
node_ptr->node_state &= NODE_STATE_FLAGS;
node_ptr->node_state |= NODE_STATE_MIXED;
}
}
}
extern int clusteracct_storage_g_node_up(void *db_conn,
struct node_record *node_ptr,
time_t event_time)
{
if (slurm_acct_storage_init(NULL) < 0)
return SLURM_ERROR;
/* on some systems we need to make sure we don't say something
is completely up if there are cpus in an error state */
if(node_ptr->select_nodeinfo) {
uint16_t err_cpus = 0;
select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ERROR,
&err_cpus);
if(err_cpus) {
char *reason = "Setting partial node down.";
struct node_record send_node;
struct config_record config_rec;
uint16_t cpu_cnt = 0;
select_g_alter_node_cnt(SELECT_GET_NODE_CPU_CNT,
&cpu_cnt);
err_cpus *= cpu_cnt;
memset(&send_node, 0, sizeof(struct node_record));
memset(&config_rec, 0, sizeof(struct config_record));
send_node.name = node_ptr->name;
send_node.config_ptr = &config_rec;
send_node.cpus = err_cpus;
config_rec.cpus = err_cpus;
send_node.node_state = NODE_STATE_ERROR;
return (*(g_acct_storage_context->ops.node_down))
(db_conn, &send_node,
event_time, reason, slurm_get_slurm_user_id());
}
}
return (*(g_acct_storage_context->ops.node_up))
(db_conn, node_ptr, event_time);
}
static int _insert_node_ptr(List sinfo_list, uint16_t part_num,
partition_info_t *part_ptr,
node_info_t *node_ptr, uint32_t node_scaling)
{
int rc = SLURM_SUCCESS;
sinfo_data_t *sinfo_ptr = NULL;
ListIterator itr = NULL;
if (params.cluster_flags & CLUSTER_FLAG_BG) {
uint16_t error_cpus = 0;
select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ERROR,
&error_cpus);
if (error_cpus && !node_ptr->reason)
node_ptr->reason = xstrdup("Block(s) in error state");
}
itr = list_iterator_create(sinfo_list);
while ((sinfo_ptr = list_next(itr))) {
if (!_match_part_data(sinfo_ptr, part_ptr))
continue;
if (sinfo_ptr->nodes_total &&
(!_match_node_data(sinfo_ptr, node_ptr)))
continue;
_update_sinfo(sinfo_ptr, node_ptr, node_scaling);
break;
}
list_iterator_destroy(itr);
/* if no match, create new sinfo_data entry */
if (!sinfo_ptr) {
list_append(sinfo_list,
_create_sinfo(part_ptr, part_num,
node_ptr, node_scaling));
}
return rc;
}
extern int slurm_get_select_nodeinfo(dynamic_plugin_data_t *nodeinfo,
enum select_nodedata_type data_type,
enum node_states state, void *data)
{
return select_g_select_nodeinfo_get(nodeinfo, data_type, state, data);
}
/* Spawn health check function for every node that is not DOWN */
extern void run_health_check(void)
{
#ifdef HAVE_FRONT_END
front_end_record_t *front_end_ptr;
#else
struct node_record *node_ptr;
int node_test_cnt = 0, node_limit, node_states, run_cyclic;
static int base_node_loc = -1;
static time_t cycle_start_time = (time_t) 0;
#endif
int i;
char *host_str = NULL;
agent_arg_t *check_agent_args = NULL;
/* Sync plugin internal data with
* node select_nodeinfo. This is important
* after reconfig otherwise select_nodeinfo
* will not return the correct number of
* allocated cpus.
*/
select_g_select_nodeinfo_set_all();
check_agent_args = xmalloc (sizeof (agent_arg_t));
check_agent_args->msg_type = REQUEST_HEALTH_CHECK;
check_agent_args->retry = 0;
check_agent_args->hostlist = hostlist_create(NULL);
#ifdef HAVE_FRONT_END
for (i = 0, front_end_ptr = front_end_nodes;
i < front_end_node_cnt; i++, front_end_ptr++) {
if (IS_NODE_NO_RESPOND(front_end_ptr))
continue;
hostlist_push_host(check_agent_args->hostlist,
front_end_ptr->name);
check_agent_args->node_count++;
}
#else
run_cyclic = slurmctld_conf.health_check_node_state &
HEALTH_CHECK_CYCLE;
node_states = slurmctld_conf.health_check_node_state &
(~HEALTH_CHECK_CYCLE);
if (run_cyclic) {
time_t now = time(NULL);
if (cycle_start_time == (time_t) 0)
cycle_start_time = now;
else if (base_node_loc >= 0)
; /* mid-cycle */
else if (difftime(now, cycle_start_time) <
slurmctld_conf.health_check_interval) {
return; /* Wait to start next cycle */
}
cycle_start_time = now;
/* Determine how many nodes we want to test on each call of
* run_health_check() to spread out the work. */
node_limit = (node_record_count * 2) /
slurmctld_conf.health_check_interval;
node_limit = MAX(node_limit, 10);
}
if ((node_states != HEALTH_CHECK_NODE_ANY) &&
(node_states != HEALTH_CHECK_NODE_IDLE)) {
/* Update each node's alloc_cpus count */
select_g_select_nodeinfo_set_all();
}
for (i = 0; i < node_record_count; i++) {
if (run_cyclic) {
if (node_test_cnt++ >= node_limit)
break;
base_node_loc++;
if (base_node_loc >= node_record_count) {
base_node_loc = -1;
break;
}
node_ptr = node_record_table_ptr + base_node_loc;
} else {
node_ptr = node_record_table_ptr + i;
}
if (IS_NODE_NO_RESPOND(node_ptr) || IS_NODE_FUTURE(node_ptr) ||
IS_NODE_POWER_SAVE(node_ptr))
continue;
if (node_states != HEALTH_CHECK_NODE_ANY) {
uint16_t cpus_total, cpus_used = 0;
if (slurmctld_conf.fast_schedule) {
cpus_total = node_ptr->config_ptr->cpus;
} else {
cpus_total = node_ptr->cpus;
}
if (!IS_NODE_IDLE(node_ptr)) {
select_g_select_nodeinfo_get(
node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ALLOCATED,
&cpus_used);
}
/* Here the node state is inferred from
* the cpus allocated on it.
* - cpus_used == 0
* means node is idle
* - cpus_used < cpus_total
* means the node is in mixed state
* else cpus_used == cpus_total
* means the node is allocated
*/
if (cpus_used == 0) {
if (!(node_states & HEALTH_CHECK_NODE_IDLE))
continue;
if (!IS_NODE_IDLE(node_ptr))
continue;
} else if (cpus_used < cpus_total) {
if (!(node_states & HEALTH_CHECK_NODE_MIXED))
continue;
} else {
if (!(node_states & HEALTH_CHECK_NODE_ALLOC))
continue;
}
}
hostlist_push_host(check_agent_args->hostlist, node_ptr->name);
check_agent_args->node_count++;
}
if (run_cyclic && (i >= node_record_count))
base_node_loc = -1;
#endif
if (check_agent_args->node_count == 0) {
hostlist_destroy(check_agent_args->hostlist);
xfree (check_agent_args);
} else {
hostlist_uniq(check_agent_args->hostlist);
host_str = hostlist_ranged_string_xmalloc(
check_agent_args->hostlist);
debug("Spawning health check agent for %s", host_str);
xfree(host_str);
ping_begin();
agent_queue_request(check_agent_args);
}
}
static void _update_sinfo(sinfo_data_t *sinfo_ptr, node_info_t *node_ptr,
uint32_t node_scaling)
{
uint16_t base_state;
uint16_t used_cpus = 0, error_cpus = 0;
int total_cpus = 0, total_nodes = 0;
/* since node_scaling could be less here, we need to use the
* global node scaling which should never change. */
int single_node_cpus = (node_ptr->cpus / g_node_scaling);
base_state = node_ptr->node_state & NODE_STATE_BASE;
if (sinfo_ptr->nodes_total == 0) { /* first node added */
sinfo_ptr->node_state = node_ptr->node_state;
sinfo_ptr->features = node_ptr->features;
sinfo_ptr->gres = node_ptr->gres;
sinfo_ptr->reason = node_ptr->reason;
sinfo_ptr->reason_time= node_ptr->reason_time;
sinfo_ptr->reason_uid = node_ptr->reason_uid;
sinfo_ptr->min_cpus = node_ptr->cpus;
sinfo_ptr->max_cpus = node_ptr->cpus;
sinfo_ptr->min_sockets = node_ptr->sockets;
sinfo_ptr->max_sockets = node_ptr->sockets;
sinfo_ptr->min_cores = node_ptr->cores;
sinfo_ptr->max_cores = node_ptr->cores;
sinfo_ptr->min_threads = node_ptr->threads;
sinfo_ptr->max_threads = node_ptr->threads;
sinfo_ptr->min_disk = node_ptr->tmp_disk;
sinfo_ptr->max_disk = node_ptr->tmp_disk;
sinfo_ptr->min_mem = node_ptr->real_memory;
sinfo_ptr->max_mem = node_ptr->real_memory;
sinfo_ptr->min_weight = node_ptr->weight;
sinfo_ptr->max_weight = node_ptr->weight;
sinfo_ptr->min_cpu_load = node_ptr->cpu_load;
sinfo_ptr->max_cpu_load = node_ptr->cpu_load;
sinfo_ptr->max_cpus_per_node = sinfo_ptr->part_info->
max_cpus_per_node;
sinfo_ptr->version = node_ptr->version;
} else if (hostlist_find(sinfo_ptr->nodes, node_ptr->name) != -1) {
/* we already have this node in this record,
* just return, don't duplicate */
return;
} else {
if (sinfo_ptr->min_cpus > node_ptr->cpus)
sinfo_ptr->min_cpus = node_ptr->cpus;
if (sinfo_ptr->max_cpus < node_ptr->cpus)
sinfo_ptr->max_cpus = node_ptr->cpus;
if (sinfo_ptr->min_sockets > node_ptr->sockets)
sinfo_ptr->min_sockets = node_ptr->sockets;
if (sinfo_ptr->max_sockets < node_ptr->sockets)
sinfo_ptr->max_sockets = node_ptr->sockets;
if (sinfo_ptr->min_cores > node_ptr->cores)
sinfo_ptr->min_cores = node_ptr->cores;
if (sinfo_ptr->max_cores < node_ptr->cores)
sinfo_ptr->max_cores = node_ptr->cores;
if (sinfo_ptr->min_threads > node_ptr->threads)
sinfo_ptr->min_threads = node_ptr->threads;
if (sinfo_ptr->max_threads < node_ptr->threads)
sinfo_ptr->max_threads = node_ptr->threads;
if (sinfo_ptr->min_disk > node_ptr->tmp_disk)
sinfo_ptr->min_disk = node_ptr->tmp_disk;
if (sinfo_ptr->max_disk < node_ptr->tmp_disk)
sinfo_ptr->max_disk = node_ptr->tmp_disk;
if (sinfo_ptr->min_mem > node_ptr->real_memory)
sinfo_ptr->min_mem = node_ptr->real_memory;
if (sinfo_ptr->max_mem < node_ptr->real_memory)
sinfo_ptr->max_mem = node_ptr->real_memory;
if (sinfo_ptr->min_weight> node_ptr->weight)
sinfo_ptr->min_weight = node_ptr->weight;
if (sinfo_ptr->max_weight < node_ptr->weight)
sinfo_ptr->max_weight = node_ptr->weight;
if (sinfo_ptr->min_cpu_load > node_ptr->cpu_load)
sinfo_ptr->min_cpu_load = node_ptr->cpu_load;
if (sinfo_ptr->max_cpu_load < node_ptr->cpu_load)
sinfo_ptr->max_cpu_load = node_ptr->cpu_load;
}
hostlist_push_host(sinfo_ptr->nodes, node_ptr->name);
if (params.match_flags.node_addr_flag)
hostlist_push_host(sinfo_ptr->node_addr, node_ptr->node_addr);
if (params.match_flags.hostnames_flag)
hostlist_push_host(sinfo_ptr->hostnames, node_ptr->node_hostname);
total_cpus = node_ptr->cpus;
total_nodes = node_scaling;
select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ALLOCATED,
&used_cpus);
select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ERROR,
&error_cpus);
if (params.cluster_flags & CLUSTER_FLAG_BG) {
if (!params.match_flags.state_flag &&
(used_cpus || error_cpus)) {
/* We only get one shot at this (because all states
* are combined together), so we need to make
* sure we get all the subgrps accounted. (So use
* g_node_scaling for safe measure) */
total_nodes = g_node_scaling;
sinfo_ptr->nodes_alloc += used_cpus;
sinfo_ptr->nodes_other += error_cpus;
sinfo_ptr->nodes_idle +=
(total_nodes - (used_cpus + error_cpus));
used_cpus *= single_node_cpus;
error_cpus *= single_node_cpus;
} else {
/* process only for this subgrp and then return */
total_cpus = total_nodes * single_node_cpus;
if ((base_state == NODE_STATE_ALLOCATED) ||
(base_state == NODE_STATE_MIXED) ||
(node_ptr->node_state & NODE_STATE_COMPLETING)) {
sinfo_ptr->nodes_alloc += total_nodes;
sinfo_ptr->cpus_alloc += total_cpus;
} else if (IS_NODE_DRAIN(node_ptr) ||
(base_state == NODE_STATE_DOWN)) {
sinfo_ptr->nodes_other += total_nodes;
sinfo_ptr->cpus_other += total_cpus;
} else {
sinfo_ptr->nodes_idle += total_nodes;
sinfo_ptr->cpus_idle += total_cpus;
}
sinfo_ptr->nodes_total += total_nodes;
sinfo_ptr->cpus_total += total_cpus;
return;
}
} else {
if ((base_state == NODE_STATE_ALLOCATED) ||
(base_state == NODE_STATE_MIXED) ||
IS_NODE_COMPLETING(node_ptr))
sinfo_ptr->nodes_alloc += total_nodes;
else if (IS_NODE_DRAIN(node_ptr)
|| (base_state == NODE_STATE_DOWN))
sinfo_ptr->nodes_other += total_nodes;
else
sinfo_ptr->nodes_idle += total_nodes;
}
sinfo_ptr->nodes_total += total_nodes;
sinfo_ptr->cpus_alloc += used_cpus;
sinfo_ptr->cpus_total += total_cpus;
total_cpus -= used_cpus + error_cpus;
if (error_cpus) {
sinfo_ptr->cpus_idle += total_cpus;
sinfo_ptr->cpus_other += error_cpus;
} else if (IS_NODE_DRAIN(node_ptr) ||
(base_state == NODE_STATE_DOWN)) {
sinfo_ptr->cpus_other += total_cpus;
} else
sinfo_ptr->cpus_idle += total_cpus;
}
/*
* _build_sinfo_data - make a sinfo_data entry for each unique node
* configuration and add it to the sinfo_list for later printing.
* sinfo_list IN/OUT - list of unique sinfo_data records to report
* partition_msg IN - partition info message
* node_msg IN - node info message
* RET zero or error code
*/
static int _build_sinfo_data(List sinfo_list,
partition_info_msg_t *partition_msg,
node_info_msg_t *node_msg)
{
pthread_attr_t attr_sinfo;
pthread_t thread_sinfo;
build_part_info_t *build_struct_ptr;
node_info_t *node_ptr = NULL;
partition_info_t *part_ptr = NULL;
int j;
g_node_scaling = node_msg->node_scaling;
/* by default every partition is shown, even if no nodes */
if ((!params.node_flag) && params.match_flags.partition_flag) {
part_ptr = partition_msg->partition_array;
for (j=0; j<partition_msg->record_count; j++, part_ptr++) {
if ((!params.partition) ||
(_strcmp(params.partition, part_ptr->name) == 0)) {
list_append(sinfo_list, _create_sinfo(
part_ptr, (uint16_t) j,
NULL,
node_msg->node_scaling));
}
}
}
if (params.filtering) {
for (j = 0; j < node_msg->record_count; j++) {
node_ptr = &(node_msg->node_array[j]);
if (node_ptr->name && _filter_out(node_ptr))
xfree(node_ptr->name);
}
}
/* make sinfo_list entries for every node in every partition */
for (j=0; j<partition_msg->record_count; j++, part_ptr++) {
part_ptr = &(partition_msg->partition_array[j]);
if (params.filtering && params.partition &&
_strcmp(part_ptr->name, params.partition))
continue;
if (node_msg->record_count == 1) { /* node_name_single */
int pos = -1;
uint16_t subgrp_size = 0;
hostlist_t hl;
node_ptr = &(node_msg->node_array[0]);
if ((node_ptr->name == NULL) ||
(part_ptr->nodes == NULL))
continue;
hl = hostlist_create(part_ptr->nodes);
pos = hostlist_find(hl, node_msg->node_array[0].name);
hostlist_destroy(hl);
if (pos < 0)
continue;
if (select_g_select_nodeinfo_get(
node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBGRP_SIZE,
0,
&subgrp_size) == SLURM_SUCCESS
&& subgrp_size) {
_handle_subgrps(sinfo_list,
(uint16_t) j,
part_ptr,
node_ptr,
node_msg->
node_scaling);
} else {
_insert_node_ptr(sinfo_list,
(uint16_t) j,
part_ptr,
node_ptr,
node_msg->
node_scaling);
}
continue;
}
/* Process each partition using a separate thread */
build_struct_ptr = xmalloc(sizeof(build_part_info_t));
build_struct_ptr->node_msg = node_msg;
build_struct_ptr->part_num = (uint16_t) j;
build_struct_ptr->part_ptr = part_ptr;
build_struct_ptr->sinfo_list = sinfo_list;
slurm_mutex_lock(&sinfo_cnt_mutex);
sinfo_cnt++;
slurm_mutex_unlock(&sinfo_cnt_mutex);
slurm_attr_init(&attr_sinfo);
if (pthread_attr_setdetachstate
(&attr_sinfo, PTHREAD_CREATE_DETACHED))
error("pthread_attr_setdetachstate error %m");
while (pthread_create(&thread_sinfo, &attr_sinfo,
_build_part_info,
(void *) build_struct_ptr)) {
error("pthread_create error %m");
usleep(10000); /* sleep and retry */
}
slurm_attr_destroy(&attr_sinfo);
}
slurm_mutex_lock(&sinfo_cnt_mutex);
while (sinfo_cnt) {
pthread_cond_wait(&sinfo_cnt_cond, &sinfo_cnt_mutex);
}
slurm_mutex_unlock(&sinfo_cnt_mutex);
_sort_hostlist(sinfo_list);
return SLURM_SUCCESS;
}
/* Build information about a partition using one pthread per partition */
void *_build_part_info(void *args)
{
build_part_info_t *build_struct_ptr;
List sinfo_list;
partition_info_t *part_ptr;
node_info_msg_t *node_msg;
node_info_t *node_ptr = NULL;
uint16_t part_num;
int j = 0;
if (_serial_part_data())
slurm_mutex_lock(&sinfo_list_mutex);
build_struct_ptr = (build_part_info_t *) args;
sinfo_list = build_struct_ptr->sinfo_list;
part_num = build_struct_ptr->part_num;
part_ptr = build_struct_ptr->part_ptr;
node_msg = build_struct_ptr->node_msg;
while (part_ptr->node_inx[j] >= 0) {
int i = 0;
uint16_t subgrp_size = 0;
for (i = part_ptr->node_inx[j];
i <= part_ptr->node_inx[j+1]; i++) {
if (i >= node_msg->record_count) {
/* If info for single node name is loaded */
break;
}
node_ptr = &(node_msg->node_array[i]);
if (node_ptr->name == NULL)
continue;
if (select_g_select_nodeinfo_get(
node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBGRP_SIZE, 0,
&subgrp_size) == SLURM_SUCCESS
&& subgrp_size) {
_handle_subgrps(sinfo_list, part_num,
part_ptr, node_ptr,
node_msg->node_scaling);
} else {
_insert_node_ptr(sinfo_list, part_num,
part_ptr, node_ptr,
node_msg->node_scaling);
}
}
j += 2;
}
xfree(args);
if (_serial_part_data())
slurm_mutex_unlock(&sinfo_list_mutex);
slurm_mutex_lock(&sinfo_cnt_mutex);
if (sinfo_cnt > 0) {
sinfo_cnt--;
} else {
error("sinfo_cnt underflow");
sinfo_cnt = 0;
}
pthread_cond_broadcast(&sinfo_cnt_cond);
slurm_mutex_unlock(&sinfo_cnt_mutex);
return NULL;
}
/*
* _query_server - download the current server state
* part_pptr IN/OUT - partition information message
* node_pptr IN/OUT - node information message
* block_pptr IN/OUT - BlueGene block data
* reserv_pptr IN/OUT - reservation information message
* clear_old IN - If set, then always replace old data, needed when going
* between clusters.
* RET zero or error code
*/
static int
_query_server(partition_info_msg_t ** part_pptr,
node_info_msg_t ** node_pptr,
block_info_msg_t ** block_pptr,
reserve_info_msg_t ** reserv_pptr,
bool clear_old)
{
static partition_info_msg_t *old_part_ptr = NULL, *new_part_ptr;
static node_info_msg_t *old_node_ptr = NULL, *new_node_ptr;
static block_info_msg_t *old_bg_ptr = NULL, *new_bg_ptr;
static reserve_info_msg_t *old_resv_ptr = NULL, *new_resv_ptr;
int error_code;
uint16_t show_flags = 0;
int cc;
node_info_t *node_ptr;
if (params.all_flag)
show_flags |= SHOW_ALL;
if (old_part_ptr) {
if (clear_old)
old_part_ptr->last_update = 0;
error_code = slurm_load_partitions(old_part_ptr->last_update,
&new_part_ptr, show_flags);
if (error_code == SLURM_SUCCESS)
slurm_free_partition_info_msg(old_part_ptr);
else if (slurm_get_errno() == SLURM_NO_CHANGE_IN_DATA) {
error_code = SLURM_SUCCESS;
new_part_ptr = old_part_ptr;
}
} else {
error_code = slurm_load_partitions((time_t) NULL, &new_part_ptr,
show_flags);
}
if (error_code) {
slurm_perror("slurm_load_partitions");
return error_code;
}
old_part_ptr = new_part_ptr;
*part_pptr = new_part_ptr;
if (old_node_ptr) {
if (clear_old)
old_node_ptr->last_update = 0;
if (params.node_name_single) {
error_code = slurm_load_node_single(&new_node_ptr,
params.nodes,
show_flags);
} else {
error_code = slurm_load_node(old_node_ptr->last_update,
&new_node_ptr, show_flags);
}
if (error_code == SLURM_SUCCESS)
slurm_free_node_info_msg(old_node_ptr);
else if (slurm_get_errno() == SLURM_NO_CHANGE_IN_DATA) {
error_code = SLURM_SUCCESS;
new_node_ptr = old_node_ptr;
}
} else if (params.node_name_single) {
error_code = slurm_load_node_single(&new_node_ptr, params.nodes,
show_flags);
} else {
error_code = slurm_load_node((time_t) NULL, &new_node_ptr,
show_flags);
}
if (error_code) {
slurm_perror("slurm_load_node");
return error_code;
}
old_node_ptr = new_node_ptr;
*node_pptr = new_node_ptr;
/* Set the node state as NODE_STATE_MIXED. */
for (cc = 0; cc < new_node_ptr->record_count; cc++) {
node_ptr = &(new_node_ptr->node_array[cc]);
if (IS_NODE_DRAIN(node_ptr)) {
/* don't worry about mixed since the
* whole node is being drained. */
} else {
uint16_t alloc_cpus = 0, err_cpus = 0, idle_cpus;
int single_node_cpus =
(node_ptr->cpus / g_node_scaling);
select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ALLOCATED,
&alloc_cpus);
if (params.cluster_flags & CLUSTER_FLAG_BG) {
if (!alloc_cpus &&
(IS_NODE_ALLOCATED(node_ptr) ||
IS_NODE_COMPLETING(node_ptr)))
alloc_cpus = node_ptr->cpus;
else
alloc_cpus *= single_node_cpus;
}
idle_cpus = node_ptr->cpus - alloc_cpus;
select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ERROR,
&err_cpus);
if (params.cluster_flags & CLUSTER_FLAG_BG)
err_cpus *= single_node_cpus;
idle_cpus -= err_cpus;
if ((alloc_cpus && err_cpus) ||
(idle_cpus && (idle_cpus != node_ptr->cpus))) {
node_ptr->node_state &= NODE_STATE_FLAGS;
node_ptr->node_state |= NODE_STATE_MIXED;
}
}
}
if (old_resv_ptr) {
if (clear_old)
old_resv_ptr->last_update = 0;
error_code = slurm_load_reservations(old_resv_ptr->last_update,
&new_resv_ptr);
if (error_code == SLURM_SUCCESS)
slurm_free_reservation_info_msg(old_resv_ptr);
else if (slurm_get_errno() == SLURM_NO_CHANGE_IN_DATA) {
error_code = SLURM_SUCCESS;
new_resv_ptr = old_resv_ptr;
}
} else {
error_code = slurm_load_reservations((time_t) NULL,
&new_resv_ptr);
}
if (error_code) {
slurm_perror("slurm_load_reservations");
return error_code;
}
old_resv_ptr = new_resv_ptr;
*reserv_pptr = new_resv_ptr;
if (!params.bg_flag)
return SLURM_SUCCESS;
if (params.cluster_flags & CLUSTER_FLAG_BG) {
if (old_bg_ptr) {
if (clear_old)
old_bg_ptr->last_update = 0;
error_code = slurm_load_block_info(
old_bg_ptr->last_update,
&new_bg_ptr, show_flags);
if (error_code == SLURM_SUCCESS)
slurm_free_block_info_msg(old_bg_ptr);
else if (slurm_get_errno() == SLURM_NO_CHANGE_IN_DATA) {
error_code = SLURM_SUCCESS;
new_bg_ptr = old_bg_ptr;
}
} else {
error_code = slurm_load_block_info((time_t) NULL,
&new_bg_ptr,
show_flags);
}
}
if (error_code) {
slurm_perror("slurm_load_block");
return error_code;
}
old_bg_ptr = new_bg_ptr;
*block_pptr = new_bg_ptr;
return SLURM_SUCCESS;
}
static int _handle_subgrps(List sinfo_list, uint16_t part_num,
partition_info_t *part_ptr,
node_info_t *node_ptr, uint32_t node_scaling)
{
uint16_t size;
int *node_state;
int i=0, state_cnt = 2;
ListIterator iterator = NULL;
enum node_states state[] =
{ NODE_STATE_ALLOCATED, NODE_STATE_ERROR };
/* If we ever update the hostlist stuff to support this stuff
* then we can use this to tack on the end of the node name
* the subgrp stuff. On bluegene systems this would be nice
* to see the ionodes in certain states.
* When asking for nodes that are reserved, we need to return
* all states of those nodes.
*/
if (params.state_list)
iterator = list_iterator_create(params.state_list);
for(i=0; i<state_cnt; i++) {
if (iterator) {
node_info_t tmp_node, *tmp_node_ptr = &tmp_node;
while ((node_state = list_next(iterator))) {
tmp_node_ptr->node_state = *node_state;
if ((((state[i] == NODE_STATE_ALLOCATED)
&& IS_NODE_DRAINING(tmp_node_ptr))
|| (*node_state == NODE_STATE_DRAIN))
|| (*node_state == state[i])
|| (*node_state == NODE_STATE_RES))
break;
}
list_iterator_reset(iterator);
if (!node_state)
continue;
}
if (select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
state[i],
&size) == SLURM_SUCCESS
&& size) {
node_scaling -= size;
node_ptr->node_state &= NODE_STATE_FLAGS;
node_ptr->node_state |= state[i];
_insert_node_ptr(sinfo_list, part_num, part_ptr,
node_ptr, size);
}
}
/* now handle the idle */
if (iterator) {
while ((node_state = list_next(iterator))) {
node_info_t tmp_node, *tmp_node_ptr = &tmp_node;
tmp_node_ptr->node_state = *node_state;
if (((*node_state == NODE_STATE_DRAIN)
|| IS_NODE_DRAINED(tmp_node_ptr))
|| (*node_state == NODE_STATE_IDLE)
|| (*node_state == NODE_STATE_RES))
break;
}
list_iterator_destroy(iterator);
if (!node_state)
return SLURM_SUCCESS;
}
node_ptr->node_state &= NODE_STATE_FLAGS;
node_ptr->node_state |= NODE_STATE_IDLE;
if ((int)node_scaling > 0)
_insert_node_ptr(sinfo_list, part_num, part_ptr,
node_ptr, node_scaling);
return SLURM_SUCCESS;
}
/*
* slurm_sprint_node_table - output information about a specific Slurm nodes
* based upon message as loaded using slurm_load_node
* IN node_ptr - an individual node information record pointer
* IN node_scaling - number of nodes each node represents
* IN one_liner - print as a single line if true
* RET out - char * containing formatted output (must be freed after call)
* NULL is returned on failure.
*/
char *
slurm_sprint_node_table (node_info_t * node_ptr,
int node_scaling, int one_liner )
{
uint32_t my_state = node_ptr->node_state;
char *cloud_str = "", *comp_str = "", *drain_str = "", *power_str = "";
char time_str[32];
char *out = NULL, *reason_str = NULL, *select_reason_str = NULL;
uint16_t err_cpus = 0, alloc_cpus = 0;
int cpus_per_node = 1;
int idle_cpus;
uint32_t cluster_flags = slurmdb_setup_cluster_flags();
uint64_t alloc_memory;
char *node_alloc_tres = NULL;
char *line_end = (one_liner) ? " " : "\n ";
if (node_scaling)
cpus_per_node = node_ptr->cpus / node_scaling;
if (my_state & NODE_STATE_CLOUD) {
my_state &= (~NODE_STATE_CLOUD);
cloud_str = "+CLOUD";
}
if (my_state & NODE_STATE_COMPLETING) {
my_state &= (~NODE_STATE_COMPLETING);
comp_str = "+COMPLETING";
}
if (my_state & NODE_STATE_DRAIN) {
my_state &= (~NODE_STATE_DRAIN);
drain_str = "+DRAIN";
}
if (my_state & NODE_STATE_FAIL) {
my_state &= (~NODE_STATE_FAIL);
drain_str = "+FAIL";
}
if (my_state & NODE_STATE_POWER_SAVE) {
my_state &= (~NODE_STATE_POWER_SAVE);
power_str = "+POWER";
}
slurm_get_select_nodeinfo(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ALLOCATED,
&alloc_cpus);
if (cluster_flags & CLUSTER_FLAG_BG) {
if (!alloc_cpus &&
(IS_NODE_ALLOCATED(node_ptr) ||
IS_NODE_COMPLETING(node_ptr)))
alloc_cpus = node_ptr->cpus;
else
alloc_cpus *= cpus_per_node;
}
idle_cpus = node_ptr->cpus - alloc_cpus;
slurm_get_select_nodeinfo(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ERROR,
&err_cpus);
if (cluster_flags & CLUSTER_FLAG_BG)
err_cpus *= cpus_per_node;
idle_cpus -= err_cpus;
if ((alloc_cpus && err_cpus) ||
(idle_cpus && (idle_cpus != node_ptr->cpus))) {
my_state &= NODE_STATE_FLAGS;
my_state |= NODE_STATE_MIXED;
}
/****** Line 1 ******/
xstrfmtcat(out, "NodeName=%s ", node_ptr->name);
if (cluster_flags & CLUSTER_FLAG_BG) {
slurm_get_select_nodeinfo(node_ptr->select_nodeinfo,
SELECT_NODEDATA_RACK_MP,
0, &select_reason_str);
if (select_reason_str) {
xstrfmtcat(out, "RackMidplane=%s ", select_reason_str);
xfree(select_reason_str);
}
}
if (node_ptr->arch)
xstrfmtcat(out, "Arch=%s ", node_ptr->arch);
xstrfmtcat(out, "CoresPerSocket=%u", node_ptr->cores);
xstrcat(out, line_end);
/****** Line ******/
xstrfmtcat(out, "CPUAlloc=%u CPUErr=%u CPUTot=%u ",
alloc_cpus, err_cpus, node_ptr->cpus);
if (node_ptr->cpu_load == NO_VAL)
xstrcat(out, "CPULoad=N/A");
else
xstrfmtcat(out, "CPULoad=%.2f", (node_ptr->cpu_load / 100.0));
xstrcat(out, line_end);
/****** Line ******/
xstrfmtcat(out, "AvailableFeatures=%s", node_ptr->features);
xstrcat(out, line_end);
/****** Line ******/
xstrfmtcat(out, "ActiveFeatures=%s", node_ptr->features_act);
xstrcat(out, line_end);
/****** Line ******/
xstrfmtcat(out, "Gres=%s", node_ptr->gres);
xstrcat(out, line_end);
/****** Line (optional) ******/
if (node_ptr->gres_drain) {
xstrfmtcat(out, "GresDrain=%s", node_ptr->gres_drain);
xstrcat(out, line_end);
}
/****** Line (optional) ******/
if (node_ptr->gres_used) {
xstrfmtcat(out, "GresUsed=%s", node_ptr->gres_used);
xstrcat(out, line_end);
}
/****** Line (optional) ******/
if (node_ptr->node_hostname || node_ptr->node_addr) {
xstrfmtcat(out, "NodeAddr=%s NodeHostName=%s Version=%s",
node_ptr->node_addr, node_ptr->node_hostname,
node_ptr->version);
xstrcat(out, line_end);
}
/****** Line ******/
if (node_ptr->os)
xstrfmtcat(out, "OS=%s ", node_ptr->os);
slurm_get_select_nodeinfo(node_ptr->select_nodeinfo,
SELECT_NODEDATA_MEM_ALLOC,
NODE_STATE_ALLOCATED,
&alloc_memory);
xstrfmtcat(out, "RealMemory=%"PRIu64" AllocMem=%"PRIu64" ",
node_ptr->real_memory, alloc_memory);
if (node_ptr->free_mem == NO_VAL64)
xstrcat(out, "FreeMem=N/A ");
else
xstrfmtcat(out, "FreeMem=%"PRIu64" ", node_ptr->free_mem);
xstrfmtcat(out, "Sockets=%u Boards=%u",
node_ptr->sockets, node_ptr->boards);
xstrcat(out, line_end);
/****** core & memory specialization Line (optional) ******/
if (node_ptr->core_spec_cnt || node_ptr->cpu_spec_list ||
node_ptr->mem_spec_limit) {
if (node_ptr->core_spec_cnt) {
xstrfmtcat(out, "CoreSpecCount=%u ",
node_ptr->core_spec_cnt);
}
if (node_ptr->cpu_spec_list) {
xstrfmtcat(out, "CPUSpecList=%s ",
node_ptr->cpu_spec_list);
}
if (node_ptr->mem_spec_limit) {
xstrfmtcat(out, "MemSpecLimit=%"PRIu64"",
node_ptr->mem_spec_limit);
}
xstrcat(out, line_end);
}
/****** Line ******/
xstrfmtcat(out, "State=%s%s%s%s%s ThreadsPerCore=%u TmpDisk=%u Weight=%u ",
node_state_string(my_state),
cloud_str, comp_str, drain_str, power_str,
node_ptr->threads, node_ptr->tmp_disk, node_ptr->weight);
if (node_ptr->owner == NO_VAL) {
xstrcat(out, "Owner=N/A ");
} else {
char *user_name = uid_to_string((uid_t) node_ptr->owner);
xstrfmtcat(out, "Owner=%s(%u) ", user_name, node_ptr->owner);
xfree(user_name);
}
xstrfmtcat(out, "MCS_label=%s",
(node_ptr->mcs_label == NULL) ? "N/A" : node_ptr->mcs_label);
xstrcat(out, line_end);
/****** Line ******/
if (node_ptr->partitions) {
xstrfmtcat(out, "Partitions=%s ", node_ptr->partitions);
xstrcat(out, line_end);
}
/****** Line ******/
if (node_ptr->boot_time) {
slurm_make_time_str((time_t *)&node_ptr->boot_time,
time_str, sizeof(time_str));
xstrfmtcat(out, "BootTime=%s ", time_str);
} else {
xstrcat(out, "BootTime=None ");
}
if (node_ptr->slurmd_start_time) {
slurm_make_time_str ((time_t *)&node_ptr->slurmd_start_time,
time_str, sizeof(time_str));
xstrfmtcat(out, "SlurmdStartTime=%s", time_str);
} else {
xstrcat(out, "SlurmdStartTime=None");
}
xstrcat(out, line_end);
/****** TRES Line ******/
select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_TRES_ALLOC_FMT_STR,
NODE_STATE_ALLOCATED, &node_alloc_tres);
xstrfmtcat(out, "CfgTRES=%s", node_ptr->tres_fmt_str);
xstrcat(out, line_end);
xstrfmtcat(out, "AllocTRES=%s",
(node_alloc_tres) ? node_alloc_tres : "");
xfree(node_alloc_tres);
xstrcat(out, line_end);
/****** Power Management Line ******/
if (!node_ptr->power || (node_ptr->power->cap_watts == NO_VAL))
xstrcat(out, "CapWatts=n/a");
else
xstrfmtcat(out, "CapWatts=%u", node_ptr->power->cap_watts);
xstrcat(out, line_end);
/****** Power Consumption Line ******/
if (!node_ptr->energy || node_ptr->energy->current_watts == NO_VAL)
xstrcat(out, "CurrentWatts=n/s LowestJoules=n/s ConsumedJoules=n/s");
else
xstrfmtcat(out, "CurrentWatts=%u "
"LowestJoules=%"PRIu64" "
"ConsumedJoules=%"PRIu64"",
node_ptr->energy->current_watts,
node_ptr->energy->base_consumed_energy,
node_ptr->energy->consumed_energy);
xstrcat(out, line_end);
/****** external sensors Line ******/
if (!node_ptr->ext_sensors
|| node_ptr->ext_sensors->consumed_energy == NO_VAL)
xstrcat(out, "ExtSensorsJoules=n/s ");
else
xstrfmtcat(out, "ExtSensorsJoules=%"PRIu64" ",
node_ptr->ext_sensors->consumed_energy);
if (!node_ptr->ext_sensors
|| node_ptr->ext_sensors->current_watts == NO_VAL)
xstrcat(out, "ExtSensorsWatts=n/s ");
else
xstrfmtcat(out, "ExtSensorsWatts=%u ",
node_ptr->ext_sensors->current_watts);
if (!node_ptr->ext_sensors
|| node_ptr->ext_sensors->temperature == NO_VAL)
xstrcat(out, "ExtSensorsTemp=n/s");
else
xstrfmtcat(out, "ExtSensorsTemp=%u",
node_ptr->ext_sensors->temperature);
xstrcat(out, line_end);
/****** Line ******/
if (node_ptr->reason && node_ptr->reason[0])
xstrcat(reason_str, node_ptr->reason);
slurm_get_select_nodeinfo(node_ptr->select_nodeinfo,
SELECT_NODEDATA_EXTRA_INFO,
0, &select_reason_str);
if (select_reason_str && select_reason_str[0]) {
if (reason_str)
xstrcat(reason_str, "\n");
xstrcat(reason_str, select_reason_str);
}
xfree(select_reason_str);
if (reason_str) {
int inx = 1;
char *save_ptr = NULL, *tok, *user_name;
tok = strtok_r(reason_str, "\n", &save_ptr);
while (tok) {
if (inx == 1) {
xstrcat(out, "Reason=");
} else {
xstrcat(out, line_end);
xstrcat(out, " ");
}
xstrfmtcat(out, "%s", tok);
if ((inx++ == 1) && node_ptr->reason_time) {
user_name = uid_to_string(node_ptr->reason_uid);
slurm_make_time_str((time_t *)&node_ptr->reason_time,
time_str, sizeof(time_str));
xstrfmtcat(out, " [%s@%s]", user_name, time_str);
xfree(user_name);
}
tok = strtok_r(NULL, "\n", &save_ptr);
}
xfree(reason_str);
}
if (one_liner)
xstrcat(out, "\n");
else
xstrcat(out, "\n\n");
return out;
}
/*
* _build_sinfo_data - make a sinfo_data entry for each unique node
* configuration and add it to the sinfo_list for later printing.
* sinfo_list IN/OUT - list of unique sinfo_data records to report
* partition_msg IN - partition info message
* node_msg IN - node info message
* RET zero or error code
*/
static int _build_sinfo_data(List sinfo_list,
partition_info_msg_t *partition_msg,
node_info_msg_t *node_msg)
{
node_info_t *node_ptr = NULL;
partition_info_t *part_ptr = NULL;
int j, j2;
g_node_scaling = node_msg->node_scaling;
/* by default every partition is shown, even if no nodes */
if ((!params.node_flag) && params.match_flags.partition_flag) {
part_ptr = partition_msg->partition_array;
for (j=0; j<partition_msg->record_count; j++, part_ptr++) {
if ((!params.partition) ||
(_strcmp(params.partition, part_ptr->name) == 0)) {
list_append(sinfo_list, _create_sinfo(
part_ptr, (uint16_t) j,
NULL,
node_msg->node_scaling));
}
}
}
/* make sinfo_list entries for every node in every partition */
for (j=0; j<partition_msg->record_count; j++, part_ptr++) {
part_ptr = &(partition_msg->partition_array[j]);
if (params.filtering && params.partition &&
_strcmp(part_ptr->name, params.partition))
continue;
j2 = 0;
while (part_ptr->node_inx[j2] >= 0) {
int i2 = 0;
uint16_t subgrp_size = 0;
for (i2 = part_ptr->node_inx[j2];
i2 <= part_ptr->node_inx[j2+1];
i2++) {
if (i2 >= node_msg->record_count) {
/* This can happen if info for single
* node name is loaded */
break;
}
node_ptr = &(node_msg->node_array[i2]);
if ((node_ptr->name == NULL) ||
(params.filtering &&
_filter_out(node_ptr)))
continue;
if (select_g_select_nodeinfo_get(
node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBGRP_SIZE,
0,
&subgrp_size) == SLURM_SUCCESS
&& subgrp_size) {
_handle_subgrps(sinfo_list,
(uint16_t) j,
part_ptr,
node_ptr,
node_msg->
node_scaling);
} else {
_insert_node_ptr(sinfo_list,
(uint16_t) j,
part_ptr,
node_ptr,
node_msg->
node_scaling);
}
}
j2 += 2;
}
}
_sort_hostlist(sinfo_list);
return SLURM_SUCCESS;
}
/*
* slurm_sprint_node_table - output information about a specific Slurm nodes
* based upon message as loaded using slurm_load_node
* IN node_ptr - an individual node information record pointer
* IN one_liner - print as a single line if true
* RET out - char * containing formatted output (must be freed after call)
* NULL is returned on failure.
*/
char *slurm_sprint_node_table(node_info_t *node_ptr, int one_liner)
{
uint32_t my_state = node_ptr->node_state;
char *cloud_str = "", *comp_str = "", *drain_str = "", *power_str = "";
char time_str[32];
char *out = NULL, *reason_str = NULL;
uint16_t alloc_cpus = 0;
int idle_cpus;
uint64_t alloc_memory;
char *node_alloc_tres = NULL;
char *line_end = (one_liner) ? " " : "\n ";
slurm_ctl_conf_info_msg_t *slurm_ctl_conf_ptr = NULL;
if (slurm_load_ctl_conf((time_t) NULL, &slurm_ctl_conf_ptr)
!= SLURM_SUCCESS)
fatal("Cannot load slurmctld conf file");
if (my_state & NODE_STATE_CLOUD) {
my_state &= (~NODE_STATE_CLOUD);
cloud_str = "+CLOUD";
}
if (my_state & NODE_STATE_COMPLETING) {
my_state &= (~NODE_STATE_COMPLETING);
comp_str = "+COMPLETING";
}
if (my_state & NODE_STATE_DRAIN) {
my_state &= (~NODE_STATE_DRAIN);
drain_str = "+DRAIN";
}
if (my_state & NODE_STATE_FAIL) {
my_state &= (~NODE_STATE_FAIL);
drain_str = "+FAIL";
}
if (my_state & NODE_STATE_POWER_SAVE) {
my_state &= (~NODE_STATE_POWER_SAVE);
power_str = "+POWER";
}
if (my_state & NODE_STATE_POWERING_DOWN) {
my_state &= (~NODE_STATE_POWERING_DOWN);
power_str = "+POWERING_DOWN";
}
slurm_get_select_nodeinfo(node_ptr->select_nodeinfo,
SELECT_NODEDATA_SUBCNT,
NODE_STATE_ALLOCATED,
&alloc_cpus);
idle_cpus = node_ptr->cpus - alloc_cpus;
if (idle_cpus && (idle_cpus != node_ptr->cpus)) {
my_state &= NODE_STATE_FLAGS;
my_state |= NODE_STATE_MIXED;
}
/****** Line 1 ******/
xstrfmtcat(out, "NodeName=%s ", node_ptr->name);
if (node_ptr->arch)
xstrfmtcat(out, "Arch=%s ", node_ptr->arch);
if (node_ptr->cpu_bind) {
char tmp_str[128];
slurm_sprint_cpu_bind_type(tmp_str, node_ptr->cpu_bind);
xstrfmtcat(out, "CpuBind=%s ", tmp_str);
}
xstrfmtcat(out, "CoresPerSocket=%u ", node_ptr->cores);
xstrcat(out, line_end);
/****** Line ******/
xstrfmtcat(out, "CPUAlloc=%u CPUTot=%u ",
alloc_cpus, node_ptr->cpus);
if (node_ptr->cpu_load == NO_VAL)
xstrcat(out, "CPULoad=N/A");
else
xstrfmtcat(out, "CPULoad=%.2f", (node_ptr->cpu_load / 100.0));
xstrcat(out, line_end);
/****** Line ******/
xstrfmtcat(out, "AvailableFeatures=%s", node_ptr->features);
xstrcat(out, line_end);
/****** Line ******/
xstrfmtcat(out, "ActiveFeatures=%s", node_ptr->features_act);
xstrcat(out, line_end);
/****** Line ******/
xstrfmtcat(out, "Gres=%s", node_ptr->gres);
xstrcat(out, line_end);
/****** Line (optional) ******/
if (node_ptr->gres_drain) {
xstrfmtcat(out, "GresDrain=%s", node_ptr->gres_drain);
xstrcat(out, line_end);
}
/****** Line (optional) ******/
if (node_ptr->gres_used) {
xstrfmtcat(out, "GresUsed=%s", node_ptr->gres_used);
xstrcat(out, line_end);
}
/****** Line (optional) ******/
{
bool line_used = false;
if (node_ptr->node_addr) {
xstrfmtcat(out, "NodeAddr=%s ", node_ptr->node_addr);
line_used = true;
}
if (node_ptr->node_hostname) {
xstrfmtcat(out, "NodeHostName=%s ",
node_ptr->node_hostname);
line_used = true;
}
if (node_ptr->port != slurm_get_slurmd_port()) {
xstrfmtcat(out, "Port=%u ", node_ptr->port);
line_used = true;
}
if (node_ptr->version &&
xstrcmp(node_ptr->version, slurm_ctl_conf_ptr->version)) {
xstrfmtcat(out, "Version=%s", node_ptr->version);
line_used = true;
}
if (line_used)
xstrcat(out, line_end);
}
/****** Line ******/
if (node_ptr->os) {
xstrfmtcat(out, "OS=%s ", node_ptr->os);
xstrcat(out, line_end);
}
/****** Line ******/
slurm_get_select_nodeinfo(node_ptr->select_nodeinfo,
SELECT_NODEDATA_MEM_ALLOC,
NODE_STATE_ALLOCATED,
&alloc_memory);
xstrfmtcat(out, "RealMemory=%"PRIu64" AllocMem=%"PRIu64" ",
node_ptr->real_memory, alloc_memory);
if (node_ptr->free_mem == NO_VAL64)
xstrcat(out, "FreeMem=N/A ");
else
xstrfmtcat(out, "FreeMem=%"PRIu64" ", node_ptr->free_mem);
xstrfmtcat(out, "Sockets=%u Boards=%u",
node_ptr->sockets, node_ptr->boards);
xstrcat(out, line_end);
/****** core & memory specialization Line (optional) ******/
if (node_ptr->core_spec_cnt || node_ptr->cpu_spec_list ||
node_ptr->mem_spec_limit) {
if (node_ptr->core_spec_cnt) {
xstrfmtcat(out, "CoreSpecCount=%u ",
node_ptr->core_spec_cnt);
}
if (node_ptr->cpu_spec_list) {
xstrfmtcat(out, "CPUSpecList=%s ",
node_ptr->cpu_spec_list);
}
if (node_ptr->mem_spec_limit) {
xstrfmtcat(out, "MemSpecLimit=%"PRIu64"",
node_ptr->mem_spec_limit);
}
xstrcat(out, line_end);
}
/****** Line ******/
xstrfmtcat(out, "State=%s%s%s%s%s ThreadsPerCore=%u TmpDisk=%u Weight=%u ",
node_state_string(my_state),
cloud_str, comp_str, drain_str, power_str,
node_ptr->threads, node_ptr->tmp_disk, node_ptr->weight);
if (node_ptr->owner == NO_VAL) {
xstrcat(out, "Owner=N/A ");
} else {
char *user_name = uid_to_string((uid_t) node_ptr->owner);
xstrfmtcat(out, "Owner=%s(%u) ", user_name, node_ptr->owner);
xfree(user_name);
}
xstrfmtcat(out, "MCS_label=%s",
(node_ptr->mcs_label == NULL) ? "N/A" : node_ptr->mcs_label);
xstrcat(out, line_end);
/****** Line ******/
if ((node_ptr->next_state != NO_VAL) &&
(my_state & NODE_STATE_REBOOT)) {
xstrfmtcat(out, "NextState=%s",
node_state_string(node_ptr->next_state));
xstrcat(out, line_end);
}
/****** Line ******/
if (node_ptr->partitions) {
xstrfmtcat(out, "Partitions=%s ", node_ptr->partitions);
xstrcat(out, line_end);
}
/****** Line ******/
if (node_ptr->boot_time) {
slurm_make_time_str((time_t *)&node_ptr->boot_time,
time_str, sizeof(time_str));
xstrfmtcat(out, "BootTime=%s ", time_str);
} else {
xstrcat(out, "BootTime=None ");
}
if (node_ptr->slurmd_start_time) {
slurm_make_time_str ((time_t *)&node_ptr->slurmd_start_time,
time_str, sizeof(time_str));
xstrfmtcat(out, "SlurmdStartTime=%s", time_str);
} else {
xstrcat(out, "SlurmdStartTime=None");
}
xstrcat(out, line_end);
/****** TRES Line ******/
select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_TRES_ALLOC_FMT_STR,
NODE_STATE_ALLOCATED, &node_alloc_tres);
xstrfmtcat(out, "CfgTRES=%s", node_ptr->tres_fmt_str);
xstrcat(out, line_end);
xstrfmtcat(out, "AllocTRES=%s",
(node_alloc_tres) ? node_alloc_tres : "");
xfree(node_alloc_tres);
xstrcat(out, line_end);
/****** Power Management Line ******/
if (!node_ptr->power || (node_ptr->power->cap_watts == NO_VAL))
xstrcat(out, "CapWatts=n/a");
else
xstrfmtcat(out, "CapWatts=%u", node_ptr->power->cap_watts);
xstrcat(out, line_end);
/****** Power Consumption Line ******/
if (!node_ptr->energy || node_ptr->energy->current_watts == NO_VAL)
xstrcat(out, "CurrentWatts=n/s AveWatts=n/s");
else
xstrfmtcat(out, "CurrentWatts=%u AveWatts=%u",
node_ptr->energy->current_watts,
node_ptr->energy->ave_watts);
xstrcat(out, line_end);
/****** external sensors Line ******/
if (!node_ptr->ext_sensors
|| node_ptr->ext_sensors->consumed_energy == NO_VAL64)
xstrcat(out, "ExtSensorsJoules=n/s ");
else
xstrfmtcat(out, "ExtSensorsJoules=%"PRIu64" ",
node_ptr->ext_sensors->consumed_energy);
if (!node_ptr->ext_sensors
|| node_ptr->ext_sensors->current_watts == NO_VAL)
xstrcat(out, "ExtSensorsWatts=n/s ");
else
xstrfmtcat(out, "ExtSensorsWatts=%u ",
node_ptr->ext_sensors->current_watts);
if (!node_ptr->ext_sensors
|| node_ptr->ext_sensors->temperature == NO_VAL)
xstrcat(out, "ExtSensorsTemp=n/s");
else
xstrfmtcat(out, "ExtSensorsTemp=%u",
node_ptr->ext_sensors->temperature);
xstrcat(out, line_end);
/****** Line ******/
if (node_ptr->reason && node_ptr->reason[0])
xstrcat(reason_str, node_ptr->reason);
if (reason_str) {
int inx = 1;
char *save_ptr = NULL, *tok, *user_name;
tok = strtok_r(reason_str, "\n", &save_ptr);
while (tok) {
if (inx == 1) {
xstrcat(out, "Reason=");
} else {
xstrcat(out, line_end);
xstrcat(out, " ");
}
xstrfmtcat(out, "%s", tok);
if ((inx++ == 1) && node_ptr->reason_time) {
user_name = uid_to_string(node_ptr->reason_uid);
slurm_make_time_str((time_t *)&node_ptr->reason_time,
time_str, sizeof(time_str));
xstrfmtcat(out, " [%s@%s]", user_name, time_str);
xfree(user_name);
}
tok = strtok_r(NULL, "\n", &save_ptr);
}
xfree(reason_str);
}
if (one_liner)
xstrcat(out, "\n");
else
xstrcat(out, "\n\n");
slurm_free_ctl_conf(slurm_ctl_conf_ptr);
return out;
}
/* Return false if this node's data needs to be added to sinfo's table of
* data to print. Return true if it is duplicate/redundant data. */
static bool _match_node_data(sinfo_data_t *sinfo_ptr, node_info_t *node_ptr)
{
uint32_t tmp = 0;
if (params.node_flag)
return false;
if (params.match_flags.hostnames_flag &&
(hostlist_find(sinfo_ptr->hostnames,
node_ptr->node_hostname) == -1))
return false;
if (params.match_flags.node_addr_flag &&
(hostlist_find(sinfo_ptr->node_addr, node_ptr->node_addr) == -1))
return false;
if (sinfo_ptr->nodes &&
params.match_flags.features_flag &&
(xstrcmp(node_ptr->features, sinfo_ptr->features)))
return false;
if (sinfo_ptr->nodes &&
params.match_flags.features_act_flag &&
(xstrcmp(node_ptr->features_act, sinfo_ptr->features_act)))
return false;
if (sinfo_ptr->nodes &&
params.match_flags.gres_flag &&
(xstrcmp(node_ptr->gres, sinfo_ptr->gres)))
return false;
if (sinfo_ptr->nodes &&
params.match_flags.reason_flag &&
(xstrcmp(node_ptr->reason, sinfo_ptr->reason)))
return false;
if (sinfo_ptr->nodes &&
params.match_flags.reason_timestamp_flag &&
(node_ptr->reason_time != sinfo_ptr->reason_time))
return false;
if (sinfo_ptr->nodes &&
params.match_flags.reason_user_flag &&
node_ptr->reason_uid != sinfo_ptr->reason_uid) {
return false;
}
if (params.match_flags.state_flag) {
char *state1, *state2;
state1 = node_state_string(node_ptr->node_state);
state2 = node_state_string(sinfo_ptr->node_state);
if (xstrcmp(state1, state2))
return false;
}
select_g_select_nodeinfo_get(node_ptr->select_nodeinfo,
SELECT_NODEDATA_MEM_ALLOC,
NODE_STATE_ALLOCATED,
&tmp);
if (params.match_flags.alloc_mem_flag &&
(tmp != sinfo_ptr->alloc_memory))
return false;
/* If no need to exactly match sizes, just return here
* otherwise check cpus, disk, memory and weigth individually */
if (!params.exact_match)
return true;
if (params.match_flags.cpus_flag &&
(node_ptr->cpus != sinfo_ptr->min_cpus))
return false;
if (params.match_flags.sockets_flag &&
(node_ptr->sockets != sinfo_ptr->min_sockets))
return false;
if (params.match_flags.cores_flag &&
(node_ptr->cores != sinfo_ptr->min_cores))
return false;
if (params.match_flags.threads_flag &&
(node_ptr->threads != sinfo_ptr->min_threads))
return false;
if (params.match_flags.sct_flag &&
((node_ptr->sockets != sinfo_ptr->min_sockets) ||
(node_ptr->cores != sinfo_ptr->min_cores) ||
(node_ptr->threads != sinfo_ptr->min_threads)))
return false;
if (params.match_flags.disk_flag &&
(node_ptr->tmp_disk != sinfo_ptr->min_disk))
return false;
if (params.match_flags.memory_flag &&
(node_ptr->real_memory != sinfo_ptr->min_mem))
return false;
if (params.match_flags.weight_flag &&
(node_ptr->weight != sinfo_ptr->min_weight))
return false;
if (params.match_flags.cpu_load_flag &&
(node_ptr->cpu_load != sinfo_ptr->min_cpu_load))
return false;
if (params.match_flags.free_mem_flag &&
(node_ptr->free_mem != sinfo_ptr->min_free_mem))
return false;
if (params.match_flags.version_flag &&
(node_ptr->version != sinfo_ptr->version))
return false;
return true;
}