burst buffer.

Bug 5665

Bug 5359

Spaces need to be built into each key definition, otherwise it'll
all run together.

Bug 5657.

%n or %h in the spool dir.

Bug 5096

Fail early when the layout requirements cannot be satisfied in
_task_layout_lllp_block().

Otherwise step layout calculations will fail, leading to a
"fatal: _task_layout_lllp_block infinite loop"

Bug 5415.

The conditional for handling --overcommit was not sufficient to prevent
other option combinations from using that code path, resulting in
unexpected resource allocation behavior. Correct that to check that
overcommit is enabled.

Bug 5345.

running.

Bug 5624

for an allocation.

Bug 5293

Backport of 9a507584 in 18.08.

Bug 5608

Tim approved

If the dbd comes up after a job array has been submitted to the
controller, the controller calls _update_job_tres() which calls
assoc_mgr_set_tres_cnt_array() which allocates memory for the job's
tres_alloc_cnt. The job array gets scheduled, but job_array_split()
doesn't NULL out the pending job's tres_alloc_cnt, so both the array
task and the pending array job are pointing to the same memory. The
array task calls job_set_alloc_tres() which free's the running job's
tres_alloc_cnt and now the pending array job is pointing to bad memory
and when the array splits again the new array task tries to free
tres_alloc_cnt in job_set_alloc_tres() and segfaults.

Bug 5604

Bug 5072

MySQL permits up to 64-character database names, but Slurm was truncating
at 33-characters. If we exceed this limit, let the mysql_query fail and
give the admin a chance to sort it out, rather than truncating and then
failing to query against the un-truncated name later on.

While here correct the fatal() message.

Bug 5586.

Bug 5554

(i.e. gres=gpu/tesla) it would get a count of 0.

Note that pipe2() is Linux-specific, but this whole cgroup plugin is
already Linux-specific in design, and the eventfd() call itself is
Linux-specific as well.

Bug 5570.

Turns out if you don't close up everything you open the hdf5 lib doesn't
like that.

Fixs regression in 011b2f23.

avoid copying strings that may be overwritten.

Bug 5547

It turns out it is fairly easy with het jobs to hit this and seg fault srun.

Bug 5390

Bug 5547.

Causes srun and sattach to fail on big endian platforms.

Bug 5556

Wrong author.

This reverts commit 392e7cfc.

a job with scontrol.

Bug 4964

not work when trying use sattach.

Bug 5556

to run the regression tests 24.* without changing that code since it links
directly to the priority plugin where that function isn't defined.

Bug 5558

Continuation of commit 21d2ab6e

Update slurm.spec and slurm.spec-legacy as well.

Fix NEWS header.

Fixes bad casting caused by Tim on review.

Make one definitive copy of sort_part_tier() in partition_msg.c and use it.

Bug 3844, 5552.

and other drain + node state flags.

Bug 5505

Bug 5505

Bug 3844.

Bug 5528.

Only split nodes here if a node_features plugin is in use.

Otherwise node fragmentation will occur if the node config has
CPUs specified but not CoresPerSocket and Sockets.

This could be avoided by filling out the node definition, but
adding this workaround for backwards compatiblity.

Bug 5039.

Task prologs could set or modify this, so wait to create the
directory until after they've finished.

Bug 5367.

There are subtle issues involved in treating a TCP transmission
as a unidirectional message delivery layer.

The original code path looks like: connect(), write(), close().
But Linux handles the write() and close() asynchronously behind the
scenes, and does not block until that write() has been ACK'd by the
remote end. So the write() and close() may succeed, even with data
still in flight. A communication error - and message loss - would
have been silently ignored, leading to unreliable message transmission.

Worse yet, one side of the connection would believe it sent the message,
while the receive side swears it never saw the packets. This leads to
infrequent and yet seemingly impossible data loss, and a very tough
bug to chase down.

This teardown code tries to force the connection to shut down in an
orderly manner, giving Slurm a chance to catch a connection problem
and the upstream calling path an opportunity to retransmit.

This teardown code is based on an approach described in Section 7.5
of "UNIX Network Programming" Volume 1 (Third Edition), specifically
the subsection regarding SO_LINGER. (And also covers why SO_LINGER is
not sufficent to prevent this issue.)

Bug 5164.

The getopt format string needs to handle an option here,
and the --help output had not been corrected after 99b2c4e8.

Bug 5522.