int main(int argc, char *argv[])
{
log_options_t opts = LOG_OPTS_STDERR_ONLY;
log_init("sbcast", opts, SYSLOG_FACILITY_DAEMON, NULL);
#ifdef HAVE_ALPS_CRAY
error("The sbcast command is not supported on Cray systems");
return 1;
#endif
#ifdef HAVE_BG
error("The sbcast command is not supported on IBM BlueGene systems");
return 1;
#endif
slurm_conf_init(NULL);
parse_command_line(argc, argv);
if (params.verbose) {
opts.stderr_level += params.verbose;
log_alter(opts, SYSLOG_FACILITY_DAEMON, NULL);
}
/* validate the source file */
if ((fd = open(params.src_fname, O_RDONLY)) < 0) {
error("Can't open `%s`: %s", params.src_fname,
strerror(errno));
exit(1);
}
if (fstat(fd, &f_stat)) {
error("Can't stat `%s`: %s", params.src_fname,
strerror(errno));
exit(1);
}
verbose("modes = %o", (unsigned int) f_stat.st_mode);
verbose("uid = %d", (int) f_stat.st_uid);
verbose("gid = %d", (int) f_stat.st_gid);
verbose("atime = %s", slurm_ctime(&f_stat.st_atime));
verbose("mtime = %s", slurm_ctime(&f_stat.st_mtime));
verbose("ctime = %s", slurm_ctime(&f_stat.st_ctime));
verbose("size = %ld", (long) f_stat.st_size);
verbose("-----------------------------");
/* identify the nodes allocated to the job */
_get_job_info();
/* transmit the file */
_bcast_file();
/* slurm_free_sbcast_cred_msg(sbcast_cred); */
exit(0);
}
/*
* scontrol_pid_info - given a local process id, print the corresponding
* slurm job id and its expected end time
* IN job_pid - the local process id of interest
*/
extern void
scontrol_pid_info(pid_t job_pid)
{
int error_code;
uint32_t job_id;
time_t end_time;
long rem_time;
error_code = slurm_pid2jobid (job_pid, &job_id);
if (error_code) {
exit_code = 1;
if (quiet_flag != 1)
slurm_perror ("slurm_pid2jobid error");
return;
}
error_code = slurm_get_end_time(job_id, &end_time);
if (error_code) {
exit_code = 1;
if (quiet_flag != 1)
slurm_perror ("slurm_get_end_time error");
return;
}
printf("Slurm job id %u ends at %s\n", job_id, slurm_ctime(&end_time));
rem_time = slurm_get_rem_time(job_id);
printf("slurm_get_rem_time is %ld\n", rem_time);
return;
}
/*****************************************************************************
* Global Print Functions
*****************************************************************************/
void print_date(void)
{
time_t now;
now = time(NULL);
printf("%s", slurm_ctime(&now));
}
/*
* Job has been notified of it's approaching time limit.
* Job will be killed shortly after timeout.
* This RPC can arrive multiple times with the same or updated timeouts.
* FIXME: We may want to signal the job or perform other action for this.
* FIXME: How much lead time do we want for this message? Some jobs may
* require tens of minutes to gracefully terminate.
*/
static void _timeout_handler(srun_timeout_msg_t *msg)
{
static time_t last_timeout = 0;
if (msg->timeout != last_timeout) {
last_timeout = msg->timeout;
verbose("job time limit to be reached at %s",
slurm_ctime(&msg->timeout));
}
}
/* _rollup_handler - Process rollup duties */
static void *_rollup_handler(void *db_conn)
{
time_t start_time = time(NULL);
time_t next_time;
/* int sigarray[] = {SIGUSR1, 0}; */
struct tm tm;
(void) pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, NULL);
(void) pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL);
if (!localtime_r(&start_time, &tm)) {
fatal("Couldn't get localtime for rollup handler %ld",
(long)start_time);
return NULL;
}
while (1) {
if (!db_conn)
break;
/* run the roll up */
slurm_mutex_lock(&rollup_lock);
running_rollup = 1;
debug2("running rollup at %s", slurm_ctime(&start_time));
acct_storage_g_roll_usage(db_conn, 0, 0, 1);
acct_storage_g_commit(db_conn, 1);
running_rollup = 0;
slurm_mutex_unlock(&rollup_lock);
/* sleep for an hour */
tm.tm_sec = 0;
tm.tm_min = 0;
tm.tm_hour++;
tm.tm_isdst = -1;
next_time = mktime(&tm);
/* get the time now we have rolled usage */
start_time = time(NULL);
sleep((next_time-start_time));
start_time = time(NULL);
if (!localtime_r(&start_time, &tm)) {
fatal("Couldn't get localtime for rollup handler %ld",
(long)start_time);
return NULL;
}
/* Just in case some new uids were added to the system
pick them up here. */
assoc_mgr_set_missing_uids();
/* repeat ;) */
}
return NULL;
}
/* set_idbuf()
* Write in the input buffer the current time and milliseconds
* the process id and the current thread id.
*/
static void
set_idbuf(char *idbuf)
{
struct timeval now;
gettimeofday(&now, NULL);
sprintf(idbuf, "%.15s.%-6d %5d %p", slurm_ctime(&now.tv_sec) + 4,
(int)now.tv_usec, (int)getpid(), (void *)pthread_self());
}
extern void print_date(void)
{
time_t now_time = time(NULL);
if (params.commandline) {
printf("%s", slurm_ctime(&now_time));
} else {
mvwprintw(text_win, main_ycord,
main_xcord, "%s",
slurm_ctime2(&now_time));
main_ycord++;
}
}
/* set_idbuf()
* Write in the input buffer the current time and milliseconds
* the process id and the current thread id.
*/
static void
set_idbuf(char *idbuf)
{
struct timeval now;
char thread_name[NAMELEN];
int max_len = 12; /* handles current longest thread name */
gettimeofday(&now, NULL);
#if HAVE_SYS_PRCTL_H
if (prctl(PR_GET_NAME, thread_name, NULL, NULL, NULL) < 0) {
error("failed to get thread name: %m");
max_len = 0;
thread_name[0] = '\0';
}
#else
/* skip printing thread name if not available */
max_len = 0;
thread_name[0] = '\0';
#endif
sprintf(idbuf, "%.15s.%-6d %5d %-*s %p", slurm_ctime(&now.tv_sec) + 4,
(int)now.tv_usec, (int)getpid(), max_len, thread_name,
(void *)pthread_self());
}
static int _print_stats(void)
{
int i;
if (!buf) {
printf("No data available. Probably slurmctld is not working\n");
return -1;
}
printf("*******************************************************\n");
printf("sdiag output at %s", slurm_ctime(&buf->req_time));
printf("Data since %s", slurm_ctime(&buf->req_time_start));
printf("*******************************************************\n");
printf("Server thread count: %d\n", buf->server_thread_count);
printf("Agent queue size: %d\n\n", buf->agent_queue_size);
printf("Jobs submitted: %d\n", buf->jobs_submitted);
printf("Jobs started: %d\n", buf->jobs_started);
printf("Jobs completed: %d\n", buf->jobs_completed);
printf("Jobs canceled: %d\n", buf->jobs_canceled);
printf("Jobs failed: %d\n", buf->jobs_failed);
printf("\nMain schedule statistics (microseconds):\n");
printf("\tLast cycle: %u\n", buf->schedule_cycle_last);
printf("\tMax cycle: %u\n", buf->schedule_cycle_max);
printf("\tTotal cycles: %u\n", buf->schedule_cycle_counter);
if (buf->schedule_cycle_counter > 0) {
printf("\tMean cycle: %u\n",
buf->schedule_cycle_sum / buf->schedule_cycle_counter);
printf("\tMean depth cycle: %u\n",
buf->schedule_cycle_depth / buf->schedule_cycle_counter);
}
if ((buf->req_time - buf->req_time_start) > 60) {
printf("\tCycles per minute: %u\n",
(uint32_t) (buf->schedule_cycle_counter /
((buf->req_time - buf->req_time_start) / 60)));
}
printf("\tLast queue length: %u\n", buf->schedule_queue_len);
if (buf->bf_active) {
printf("\nBackfilling stats (WARNING: data obtained"
" in the middle of backfilling execution.)\n");
} else
printf("\nBackfilling stats\n");
printf("\tTotal backfilled jobs (since last slurm start): %u\n",
buf->bf_backfilled_jobs);
printf("\tTotal backfilled jobs (since last stats cycle start): %u\n",
buf->bf_last_backfilled_jobs);
printf("\tTotal cycles: %u\n", buf->bf_cycle_counter);
printf("\tLast cycle when: %s", slurm_ctime(&buf->bf_when_last_cycle));
printf("\tLast cycle: %u\n", buf->bf_cycle_last);
printf("\tMax cycle: %u\n", buf->bf_cycle_max);
if (buf->bf_cycle_counter > 0) {
printf("\tMean cycle: %"PRIu64"\n",
buf->bf_cycle_sum / buf->bf_cycle_counter);
}
printf("\tLast depth cycle: %u\n", buf->bf_last_depth);
printf("\tLast depth cycle (try sched): %u\n", buf->bf_last_depth_try);
if (buf->bf_cycle_counter > 0) {
printf("\tDepth Mean: %u\n",
buf->bf_depth_sum / buf->bf_cycle_counter);
printf("\tDepth Mean (try depth): %u\n",
buf->bf_depth_try_sum / buf->bf_cycle_counter);
}
printf("\tLast queue length: %u\n", buf->bf_queue_len);
if (buf->bf_cycle_counter > 0) {
printf("\tQueue length mean: %u\n",
buf->bf_queue_len_sum / buf->bf_cycle_counter);
}
printf("\nRemote Procedure Call statistics by message type\n");
for (i = 0; i < buf->rpc_type_size; i++) {
printf("\t%-40s(%5u) count:%-6u "
"ave_time:%-6u total_time:%"PRIu64"\n",
rpc_num2string(buf->rpc_type_id[i]),
buf->rpc_type_id[i], buf->rpc_type_cnt[i],
rpc_type_ave_time[i], buf->rpc_type_time[i]);
}
printf("\nRemote Procedure Call statistics by user\n");
for (i = 0; i < buf->rpc_user_size; i++) {
printf("\t%-16s(%8u) count:%-6u "
"ave_time:%-6u total_time:%"PRIu64"\n",
uid_to_string_cached((uid_t)buf->rpc_user_id[i]),
buf->rpc_user_id[i], buf->rpc_user_cnt[i],
rpc_user_ave_time[i], buf->rpc_user_time[i]);
}
return 0;
}
extern int as_mysql_job_start(mysql_conn_t *mysql_conn,
struct job_record *job_ptr)
{
int rc=SLURM_SUCCESS;
char *nodes = NULL, *jname = NULL, *node_inx = NULL;
int track_steps = 0;
char *block_id = NULL, *partition = NULL,
*gres_req = NULL, *gres_alloc = NULL;
char *query = NULL;
int reinit = 0;
time_t begin_time, check_time, start_time, submit_time;
uint32_t wckeyid = 0;
int job_state, node_cnt = 0;
uint32_t job_db_inx = job_ptr->db_index;
if ((!job_ptr->details || !job_ptr->details->submit_time)
&& !job_ptr->resize_time) {
error("as_mysql_job_start: "
"Not inputing this job, it has no submit time.");
return SLURM_ERROR;
}
if (check_connection(mysql_conn) != SLURM_SUCCESS)
return ESLURM_DB_CONNECTION;
debug2("as_mysql_slurmdb_job_start() called");
job_state = job_ptr->job_state;
if (job_ptr->resize_time) {
begin_time = job_ptr->resize_time;
submit_time = job_ptr->resize_time;
start_time = job_ptr->resize_time;
} else {
begin_time = job_ptr->details->begin_time;
submit_time = job_ptr->details->submit_time;
start_time = job_ptr->start_time;
}
/* Since we need a new db_inx make sure the old db_inx
* removed. This is most likely the only time we are going to
* be notified of the change also so make the state without
* the resize. */
if (IS_JOB_RESIZING(job_ptr)) {
/* If we have a db_index lets end the previous record. */
if (!job_ptr->db_index) {
error("We don't have a db_index for job %u, "
"this should only happen when resizing "
"jobs and the database interface was down.",
job_ptr->job_id);
job_ptr->db_index = _get_db_index(mysql_conn,
job_ptr->details->
submit_time,
job_ptr->job_id,
job_ptr->assoc_id);
}
if (job_ptr->db_index)
as_mysql_job_complete(mysql_conn, job_ptr);
job_state &= (~JOB_RESIZING);
job_ptr->db_index = 0;
}
job_state &= JOB_STATE_BASE;
/* See what we are hearing about here if no start time. If
* this job latest time is before the last roll up we will
* need to reset it to look at this job. */
if (start_time)
check_time = start_time;
else if (begin_time)
check_time = begin_time;
else
check_time = submit_time;
slurm_mutex_lock(&rollup_lock);
if (check_time < global_last_rollup) {
MYSQL_RES *result = NULL;
MYSQL_ROW row;
/* check to see if we are hearing about this time for the
* first time.
*/
query = xstrdup_printf("select job_db_inx "
"from \"%s_%s\" where id_job=%u and "
"time_submit=%ld and time_eligible=%ld "
"and time_start=%ld;",
mysql_conn->cluster_name,
job_table, job_ptr->job_id,
submit_time, begin_time, start_time);
debug3("%d(%s:%d) query\n%s",
mysql_conn->conn, THIS_FILE, __LINE__, query);
if (!(result =
mysql_db_query_ret(mysql_conn, query, 0))) {
xfree(query);
slurm_mutex_unlock(&rollup_lock);
return SLURM_ERROR;
}
xfree(query);
if ((row = mysql_fetch_row(result))) {
mysql_free_result(result);
debug4("revieved an update for a "
"job (%u) already known about",
job_ptr->job_id);
slurm_mutex_unlock(&rollup_lock);
goto no_rollup_change;
}
mysql_free_result(result);
if (job_ptr->start_time)
debug("Need to reroll usage from %sJob %u "
"from %s started then and we are just "
"now hearing about it.",
slurm_ctime(&check_time),
job_ptr->job_id, mysql_conn->cluster_name);
else if (begin_time)
debug("Need to reroll usage from %sJob %u "
"from %s became eligible then and we are just "
"now hearing about it.",
slurm_ctime(&check_time),
job_ptr->job_id, mysql_conn->cluster_name);
else
debug("Need to reroll usage from %sJob %u "
"from %s was submitted then and we are just "
"now hearing about it.",
slurm_ctime(&check_time),
job_ptr->job_id, mysql_conn->cluster_name);
global_last_rollup = check_time;
slurm_mutex_unlock(&rollup_lock);
/* If the times here are later than the daily_rollup
or monthly rollup it isn't a big deal since they
are always shrunk down to the beginning of each
time period.
*/
query = xstrdup_printf("update \"%s_%s\" set "
"hourly_rollup=%ld, "
"daily_rollup=%ld, monthly_rollup=%ld",
mysql_conn->cluster_name,
last_ran_table, check_time,
check_time, check_time);
debug3("%d(%s:%d) query\n%s",
mysql_conn->conn, THIS_FILE, __LINE__, query);
rc = mysql_db_query(mysql_conn, query);
xfree(query);
} else
slurm_mutex_unlock(&rollup_lock);
no_rollup_change:
if (job_ptr->name && job_ptr->name[0])
jname = slurm_add_slash_to_quotes(job_ptr->name);
else {
jname = xstrdup("allocation");
track_steps = 1;
}
if (job_ptr->nodes && job_ptr->nodes[0])
nodes = job_ptr->nodes;
else
nodes = "None assigned";
if (job_ptr->batch_flag)
track_steps = 1;
if (slurmdbd_conf) {
block_id = xstrdup(job_ptr->comment);
node_cnt = job_ptr->total_nodes;
node_inx = job_ptr->network;
} else {
char temp_bit[BUF_SIZE];
if (job_ptr->node_bitmap) {
node_inx = bit_fmt(temp_bit, sizeof(temp_bit),
job_ptr->node_bitmap);
}
#ifdef HAVE_BG
select_g_select_jobinfo_get(job_ptr->select_jobinfo,
SELECT_JOBDATA_BLOCK_ID,
&block_id);
select_g_select_jobinfo_get(job_ptr->select_jobinfo,
SELECT_JOBDATA_NODE_CNT,
&node_cnt);
#else
node_cnt = job_ptr->total_nodes;
#endif
}
/* If there is a start_time get the wckeyid. If the job is
* cancelled before the job starts we also want to grab it. */
if (job_ptr->assoc_id
&& (job_ptr->start_time || IS_JOB_CANCELLED(job_ptr)))
wckeyid = _get_wckeyid(mysql_conn, &job_ptr->wckey,
job_ptr->user_id,
mysql_conn->cluster_name,
job_ptr->assoc_id);
if (job_ptr->partition)
partition = slurm_add_slash_to_quotes(job_ptr->partition);
if (job_ptr->gres_req)
gres_req = slurm_add_slash_to_quotes(job_ptr->gres_req);
if (job_ptr->gres_alloc)
gres_alloc = slurm_add_slash_to_quotes(job_ptr->gres_alloc);
if (!job_ptr->db_index) {
if (!begin_time)
begin_time = submit_time;
query = xstrdup_printf(
"insert into \"%s_%s\" "
"(id_job, id_array_job, id_array_task, "
"id_assoc, id_qos, id_wckey, id_user, "
"id_group, nodelist, id_resv, timelimit, "
"time_eligible, time_submit, time_start, "
"job_name, track_steps, state, priority, cpus_req, "
"cpus_alloc, nodes_alloc, mem_req",
mysql_conn->cluster_name, job_table);
if (job_ptr->account)
xstrcat(query, ", account");
if (partition)
xstrcat(query, ", `partition`");
if (block_id)
xstrcat(query, ", id_block");
if (job_ptr->wckey)
xstrcat(query, ", wckey");
if (node_inx)
xstrcat(query, ", node_inx");
if (gres_req)
xstrcat(query, ", gres_req");
if (gres_alloc)
xstrcat(query, ", gres_alloc");
xstrfmtcat(query,
") values (%u, %u, %u, %u, %u, %u, %u, %u, "
"'%s', %u, %u, %ld, %ld, %ld, "
"'%s', %u, %u, %u, %u, %u, %u, %u",
job_ptr->job_id, job_ptr->array_job_id,
job_ptr->array_task_id, job_ptr->assoc_id,
job_ptr->qos_id, wckeyid,
job_ptr->user_id, job_ptr->group_id, nodes,
job_ptr->resv_id, job_ptr->time_limit,
begin_time, submit_time, start_time,
jname, track_steps, job_state,
job_ptr->priority, job_ptr->details->min_cpus,
job_ptr->total_cpus, node_cnt,
job_ptr->details->pn_min_memory);
if (job_ptr->account)
xstrfmtcat(query, ", '%s'", job_ptr->account);
if (partition)
xstrfmtcat(query, ", '%s'", partition);
if (block_id)
xstrfmtcat(query, ", '%s'", block_id);
if (job_ptr->wckey)
xstrfmtcat(query, ", '%s'", job_ptr->wckey);
if (node_inx)
xstrfmtcat(query, ", '%s'", node_inx);
if (gres_req)
xstrfmtcat(query, ", '%s'", gres_req);
if (gres_alloc)
xstrfmtcat(query, ", '%s'", gres_alloc);
xstrfmtcat(query,
") on duplicate key update "
"job_db_inx=LAST_INSERT_ID(job_db_inx), "
"id_wckey=%u, id_user=%u, id_group=%u, "
"nodelist='%s', id_resv=%u, timelimit=%u, "
"time_submit=%ld, time_start=%ld, "
"job_name='%s', track_steps=%u, id_qos=%u, "
"state=greatest(state, %u), priority=%u, "
"cpus_req=%u, cpus_alloc=%u, nodes_alloc=%u, "
"mem_req=%u, id_array_job=%u, id_array_task=%u",
wckeyid, job_ptr->user_id, job_ptr->group_id, nodes,
job_ptr->resv_id, job_ptr->time_limit,
submit_time, start_time,
jname, track_steps, job_ptr->qos_id, job_state,
job_ptr->priority, job_ptr->details->min_cpus,
job_ptr->total_cpus, node_cnt,
job_ptr->details->pn_min_memory,
job_ptr->array_job_id,
job_ptr->array_task_id);
if (job_ptr->account)
xstrfmtcat(query, ", account='%s'", job_ptr->account);
if (partition)
xstrfmtcat(query, ", `partition`='%s'", partition);
if (block_id)
xstrfmtcat(query, ", id_block='%s'", block_id);
if (job_ptr->wckey)
xstrfmtcat(query, ", wckey='%s'", job_ptr->wckey);
if (node_inx)
xstrfmtcat(query, ", node_inx='%s'", node_inx);
if (gres_req)
xstrfmtcat(query, ", gres_req='%s'", gres_req);
if (gres_alloc)
xstrfmtcat(query, ", gres_alloc='%s'", gres_alloc);
debug3("%d(%s:%d) query\n%s",
mysql_conn->conn, THIS_FILE, __LINE__, query);
try_again:
if (!(job_ptr->db_index = mysql_db_insert_ret_id(
mysql_conn, query))) {
if (!reinit) {
error("It looks like the storage has gone "
"away trying to reconnect");
mysql_db_close_db_connection(
mysql_conn);
/* reconnect */
check_connection(mysql_conn);
reinit = 1;
goto try_again;
} else
rc = SLURM_ERROR;
}
} else {
query = xstrdup_printf("update \"%s_%s\" set nodelist='%s', ",
mysql_conn->cluster_name,
job_table, nodes);
if (job_ptr->account)
xstrfmtcat(query, "account='%s', ", job_ptr->account);
if (partition)
xstrfmtcat(query, "`partition`='%s', ", partition);
if (block_id)
xstrfmtcat(query, "id_block='%s', ", block_id);
if (job_ptr->wckey)
xstrfmtcat(query, "wckey='%s', ", job_ptr->wckey);
if (node_inx)
xstrfmtcat(query, "node_inx='%s', ", node_inx);
if (gres_req)
xstrfmtcat(query, "gres_req='%s', ", gres_req);
if (gres_alloc)
xstrfmtcat(query, "gres_alloc='%s', ", gres_alloc);
xstrfmtcat(query, "time_start=%ld, job_name='%s', state=%u, "
"cpus_alloc=%u, nodes_alloc=%u, id_qos=%u, "
"id_assoc=%u, id_wckey=%u, id_resv=%u, "
"timelimit=%u, mem_req=%u, "
"id_array_job=%u, id_array_task=%u, "
"time_eligible=%ld where job_db_inx=%d",
start_time, jname, job_state,
job_ptr->total_cpus, node_cnt, job_ptr->qos_id,
job_ptr->assoc_id, wckeyid,
job_ptr->resv_id, job_ptr->time_limit,
job_ptr->details->pn_min_memory,
job_ptr->array_job_id,
job_ptr->array_task_id,
begin_time, job_ptr->db_index);
debug3("%d(%s:%d) query\n%s",
mysql_conn->conn, THIS_FILE, __LINE__, query);
rc = mysql_db_query(mysql_conn, query);
}
xfree(block_id);
xfree(partition);
xfree(gres_req);
xfree(gres_alloc);
xfree(jname);
xfree(query);
/* now we will reset all the steps */
if (IS_JOB_RESIZING(job_ptr)) {
/* FIXME : Verify this is still needed */
if (IS_JOB_SUSPENDED(job_ptr))
as_mysql_suspend(mysql_conn, job_db_inx, job_ptr);
}
return rc;
}
