this is happening because the energy status MSR registers for Package: "MSR_PKG_ENERGY_STATUS" and DRAM: "MSR_DRAM_ENERGY_STATUS"
are counted on 32 bit ( the MSRs are 64bit wide) and they wrap around every few minutes.

for info check out the attached screenshot of the Intel's doc
or here for Intel's complete doc: http://download.intel.com/products/processor/manual/325384.pdf

user mark the state canceled instead of completed.

so it gets sent again.  This isn't a major problem since the start will
happen when the job ends, but this does make things cleaner.

you query against that with -N and -E you will get all jobs during that
time instead of only the ones running on -N.

Signed-off-by: Danny Auble <da@schedmd.com>

to a list, which can be arbitrary size

was already in error and isn't deallocating and underlying hardware goes
bad one could get overlapping blocks in error making the code assert when
a new job request comes in.

overcommit.

overcommit.

Previously only requeued the job once

Without this change, the nodes associated with a reservation would
only be updated if the partition's nodes were reset using the
"scontrol update partition" command, but not if they were reset
using "scontrol reconfigure"

written by other cpus sharing the package