Since ReleaseAgent is no longer required, we can strip out
all the supporting logic for it.

Accidentally removed bracked in checked-in code

bug 3824

Do not run the Node Health Check on termination of the external
  step as this happens when the job allocation ends and the job
  NHC will be executed anyway.
Bug 4074

Reported by Clang

Prevent a heterogeneous job allocation from including the same nodes in
      multiple components (required by MPI jobs spanning components).

A lot of slurmdbd operations are authenticated by the accounting_storage
plugin, rather than in slurmdbd. To allow the drop_priv flag to work it
must be checked in is_user_min_admin_level() in addition to the various
functions in proc_req.c .

This reverts commit 47dad9e8.

On second thought, we should be using these or quite similar functions.

This reverts commit a44ef130.

A lot of slurmdbd operations are authenticated by the accounting_storage
plugin, rather than in slurmdbd. To allow the drop_priv flag to work it
must be checked in is_user_min_admin_level() in addition to the various
functions in proc_req.c .

Signed-off-by: Artem Polyakov <artpol84@gmail.com>

Add new performance debugging feature to pmix plugin that allows to
measure plain collectives performance on selected message sizes.
This functionality is similar to a ping-pong feature.

Performance results for existing point-to-point modes:

size    sapi            dtcp            ducx
1       0.002319442     0.000492334     0.000135243
2       0.002318223     0.000453552     0.000137356
4       0.00227046      0.00045832      0.000137117
8       0.002342675     0.000463539     0.000136455
16      0.00235131      0.000481208     0.00013619
32      0.002333058     0.000562986     0.000140756
64      0.002456691     0.000883791     0.000142574
128     0.002953556     0.001326429     0.000142336
256     0.003892236     0.002324766     0.000161224
512     0.006044123     0.004371988     0.000177675
1024    0.010324001     0.008485476     0.000224325
2048    0.018556118     0.016488896     0.000347243
4096    0.035331223     0.032744778     0.000481764
8192    0.06957123      0.065519465     0.001194106
16384   0.137925333     0.130130662     0.002544668
32768   0.272100422     0.259290563     0.009916888
65536   0.543431362     0.486692217     0.012841119

Signed-off-by: Artem Polyakov <artpol84@gmail.com>

Signed-off-by: Artem Polyakov <artpol84@gmail.com>

Signed-off-by: Artem Polyakov <artpol84@gmail.com>

Signed-off-by: Artem Polyakov <artpol84@gmail.com>

There were segmentation faults because of double free of a pending
list when UCX comonent was trying to connect multiple times.

Signed-off-by: Artem Polyakov <artpol84@gmail.com>

Statically linked Cray PMI applications still expect to use some file paths
containing the old SLURM_ID_HASH format. Some Cray customers have
certification requirements that make recompilation difficult.

The attached patch defines a macro to convert the new SLURM_ID_HASH
to the old format, and writes the files and symlinks necessary for statically
linked Cray PMI applications to work.

Bug 4114

Bug 4090

This applies to job steps for MPI. Even if there is more than one
  pack job component in a single MPI_COMM_WORLD, they will share
  a common SLURM_JOBID.

as reported when compiling with optimizations (-O2).

as reported when compiling with optimizations (-02). Initialize the
variables early since they were being initialized in their loops and
later checked for -1.

Technically couldn't have happened since for example, user_part_inx1
would only be set to -1 if max_backfill_job_per_user_part was set. And
user_part_inx is only checked later if max_backfill_job_per_user_part
is set.

Same thing for part_inx, max_backfill_job_per_part and user_inx,
max_backfill_job_per_user.

reported when compiling with optimizations (-O2). The compiler ignores
the (void) cast and reports the error.

reported when compiling with optimizations (-O2). field_id may be
uninitialized or the the value from the previous iteration in the while
loop. The only possible values of dataset_loc->type are:

typedef enum {
	PROFILE_FIELD_NOT_SET,
	PROFILE_FIELD_UINT64,
	PROFILE_FIELD_DOUBLE
} acct_gather_profile_field_type_t;

and the while loop condition ensures that PROFILE_FIELD_NOT_SET is not
handled. So instead of handling PROFILE_FIELD_NOT_SET directly, just
catching everything else with the "default" case statement and
continuing the loop.

relies on the primary to do so.

There is a potential race condition if the backup DBD tries to create/check the
database at the same time as the primary.  This patch removes this race by not
allowing the backup to do the check/create.

Bug 3827

Coverity 44943, 44944

Coverity CID 44941, 44942

This sets more per-pack-job environment variables for launched steps.
All of the following are used by Open MPI:
SLURM_CPUS_PER_TASK
SLURM_STEP_NUM_TASKS
SLURM_TASKS_PER_NODE
A few more env vars are still needed by OpenMPI

These are required by OpenMPI

Coverity CID 44723

Testing if curl_handle != NULL or rc != SLURM_SUCCESS was already done
in the right above if/else statements, jumping to the consequent goto
cleanup label if needed. Thus the removed test was never going to be
evaluated to true, and Coverity properly warned about this.

Regression introduced in commit 5f5e6472 (code cleanup).

Running slurmctld under valgrind while operating with jobcomp/elasticsearch
reported the following bytes definitely lost:

==27403== 658 bytes in 1 blocks are definitely lost in loss record 301 of 342
==27403==    at 0x4C2FD4F: realloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==27403==    by 0x2281B3: slurm_xrealloc (xmalloc.c:137)
==27403==    by 0x22856A: makespace (xstring.c:114)
==27403==    by 0x2285D0: _xstrcat (xstring.c:132)
==27403==    by 0x228CE0: _xstrfmtcat (xstring.c:291)
==27403==    by 0x83C5BCD: ???
==27403==    by 0x30A913: g_slurm_jobcomp_write (slurm_jobcomp.c:172)
==27403==    by 0x18D8FC: job_completion_logger (job_mgr.c:13652)

It turns out the generated buffer in slurm_jobcomp_log_record was xstrdup'ed to
the corresponding job_node->serialized_job, but the originally generated buffer
wasn't freed afterwards. The fix consists in change the transfer so that instead
of xstrdup'ing the char * we just assign the pointer and NULL the buffer.

The job_node->serialized_job was already xfree'd properly later when the job
was indexed.

Discovered while working on Bug 4065.

This should only happen due to ESLURM_RESULT_TOO_LARGE,
which leads to no list being packed.

Follow on to 390da8cf / 8cf1835c.

Bug 3624.