Since we don't currently track energy usage per task (only per step).
Otherwise we get double the energy.

acctg-freq is a string instead of a uint16_t

acct_gather_infiniband_g_node_init

are called.

If the slurmstepd connects task I/O, but aborts after srun accepts the connect
and before slurmstepd writes data then srun could possibly hand indefinitely.
This probably does not explain failures seen at CEA, but can't hurt matters.
then the sr

A dedicated thread (_kill_thr) is launched by slurmstepd at the end of a
step in order to destroy the IO thread if it does not manage to correctly
terminate by itself after 300 seconds.

Two bugs are corrected in this logic by this patch.

First, the performed sleep(300) is not protected against interruptions
and this delay can be reduced to a few seconds in case of signals received
by slurmstepd, thus, reducing the delay and forcing the IO thread to
terminate before the expiration of the grace time. The logic is modified
to ensure that the delay is respected using a loop around the sleep().

Second, to terminate the IO thread, a SIGKILL is delivered to the IO thread
using pthread_kill. However, sending SIGKILL using pthread_kill is a
process-wide operation (see man pthread_kill), thus all the slurmstepd
threads are killed and slurmstepd is terminated. This logic is modified
by using pthread_cancel() instead of pthread_kill() thus letting the
pthread_join() of _wait_for_io() having a chance to act as expected.

Without this patch, when _kill_thr is interrupted, slurmstepd is
terminated, letting the step in a incomplete state, as the node may not
have been able to send the REQUEST_STEP_COMPLETE to the controler.
Thus, consecutive steps can no longer be executed and stay permanently in
the "Job step creation temporarily disabled, retrying" state.

the attached patch changes the default timestamp format in logfiles to conform to RFC 5424 (the current version of the syslog RFC). It is identical to the current default "ISO 8601" timestamp used by slurm, with the exception that the timezone offset is appended. This has the benefits of

1) It's unambiguous.

2) Avoids potential confusion for admins running cluster(s) in different timezones.

3) Might help debug issues related to DST transitions. (More on that later..)

(To be pedantic, a RFC 5424 timestamp is still a valid ISO 8601 timestamp, but the converse is not necessarily true. So there is RFC 3339 which is a "profile" of ISO 8601, that is a subset, recommended for internet protocols. The RFC 5424 timestamp, in turn, is a subset of the RFC 3339 timestamps.)

The previous behavior of can be used by running configure with the

--disable-rfc5424time

flag.

This privileged call is executed just after the fork of each
forked task by user root before becoming the user. It enables
the task plugin to perform actions as a privileged user in
every task context.

provides the following improvements on the energy accounting framework :
1)the per step average frequency is now calculated correctly and may be reported by sstat and sacct
2)correction on the logic of per step energy consumption calculation through rapl plugin. This value may be also reported through sstat and sacct
3)node power and energy monitoring now working correctly through rapl plugin.

If exec_wait_child_wait_for_parent() fails for any reason, it is safer
to abort immediately rather than proceed to execute the user's job.

On a failure of fork(2), slurmstepd would print an error and exit,
possibly leaving previously forked children waiting.

Ensure a better cleanup by killing all active children on fork failure
before exiting slurmstepd.

Close the read end of the pipe slurmstepd uses to notify children
it is time to call exec(2) in order to save one file descriptor per
task. (Previously, the read side of the pipe wasn't closed until
exec_wait_info was destroyed)

maintain.

Original patch from Martin Perry (Bull)

This will eventually permit the slurmctld to respond to the slurmd
with a new task launch RPC

Cray - Add support for zero compute note resource allocation to run batch
script on front-end node with no ALPS reservation. Useful for pre- or post-
processing. NOTE: The partition must be configured with MinNodes=0.

Add job name to call. Add logging for call.

Access to secured FS often requires to have a valid token in the
user context. With SLURM, this token can be obtained using one of
the possible pluggable architecture, SPANK or PAM.

IO setup of SLURM can require to access secured FS (stdout/stderr
files). This patch ensures that pluggable frameworks are activated
and called prior to IO setup and that IO are terminated before
calling pluggable framework exit calls.

set PR_DUMPABLE as soon as possible, especially before any plugins
are loaded. This will allow someone debugging to get a coredump.

To prepare io_setup integration in _fork_all_tasks, error handling
must be transformed to not always return SLURM_ERROR but be prepared
to return SLURM_SUCCESS in case of an io_setup error.

Add call to mpi_hook_slurmstepd_prefork() from slurmstep
immediately prior to fork/exec of user tasks.
Patch from Hongjia Cao, NUDT.

It was found that slurmstepd was intermittently leaving SIGPIPE
blocked when launching user tasks. This may have something to do
with the fact that the xsignal_unblock() call in _fork_all_tasks()
is referencing an extern array (nominally this should have unblocked
SIGPIPE), but I didn't spend the time to fully track this issue
down. Instead, I figured there is probably no reason we would _not_
want to unblock *all* signals, so this patch does that.

Before this change, the following program fails every once in awhile:

 #include <stdio.h>
 #include <signal.h>

 int main (int ac, char **av)
 {
	int i, rc = 0;
	struct sigaction act;
	for (i = 1; i < SIGRTMAX; i++) {
		sigaction (i, NULL, &act);
		if (act.sa_handler == SIG_DFL)
			continue;
		fprintf (stderr, "Signal %d appears to be ignored!\n", i);
		rc = 1;
	}
	return (rc);
 }

with:

 srun -N1 -n1 ./test
 Signal 13 appears to be ignored!

after the change, the program succeeds.