[HTCondor-users] bug in schedd_negotiate.cpp

[Thomas Hartmann <thomas.hartmann@xxxxxxxx>]

hi,
as nobody has so far replied to my other post on this mailing list or to 
the mail that i sent directly to the htcondor-admin address, i try it 
again here:

there is a serious bug in schedd_negotiate.cpp, line 267:

disk = (MAX((int) ceil((disk / total_disk) * 100), 1)) * 
int(total_disk/100.0);

this line calculates the disk space that is allocated to a slot. the 
problem is that if a job requests less than 1% of the total available 
disk space, this is ignored. instead, it is treated as having requested 
this 1% of the disk space.

in normal systems with a small amount of CPU cores, this will not really 
be a problem. however, i have an execution machine with 144 cores. as 
the minimum disk space a slot gets is 1% of the available disk space, 
the scheduler runs out of disk space after 100 jobs.

i have not found a bug tracking system to which i could post this. 
neither is there an issue tracker on github. i would be very happy to 
help to resolve this bug.

best,
thomas

--
Dr. Thomas Hartmann

Centre for Cognitive Neuroscience
FB Psychologie
UniversitÃt Salzburg
HellbrunnerstraÃe 34/II
5020 Salzburg

Tel: +43 662 8044 5109
Email: thomas.hartmann@xxxxxxxx

"I am a brain, Watson. The rest of me is a mere appendix. " (Arthur Conan Doyle)