genome / gms

The Genome Modeling System installer
https://github.com/genome/gms/wiki
GNU Lesser General Public License v3.0
78 stars 22 forks source link

Jobs submitted to openlava report having insufficient slots... #10

Closed malachig closed 11 years ago

malachig commented 11 years ago

Alignment jobs immediately enter a 'pending' state. bjobs -l reports the following:

PENDING REASONS: Not enough job slot(s); Not enough job slot(s);

The resource request looks like this:

4 Processors Requested, Requested Resources <select[ncpus>=4 && mem>=8192] span[hosts=1] rusage[mem=8192]>

It seems that if you do 'bmod -n 1', the number of 'Processors Requested' is reduced and the job seems to fire off. Changing the 'Requested Resources' with 'bmod -R ...' does not seem to have an effect as long as 'Processors Requested' is greater than 1.

I am testing on a VM that thinks it has 24 cpus. Not sure if this problem is VM specific or just related to openlava...

sakoht commented 11 years ago

I just pushed a change to not use his with openlava. It seems redundant with select and rusage. If it is not, restoring its functionality may improve efficient use of resources with openlava.

On Sep 30, 2013, at 6:10 PM, Malachi Griffith notifications@github.com wrote:

Alignment jobs immediately enter a 'pending' state. bjobs -l reports the following:

PENDING REASONS: Not enough job slot(s); Not enough job slot(s);

The resource request looks like this:

4 Processors Requested, Requested Resources =4 && mem>=8192] span[hosts=1] rusage[mem=8192]>

It seems that if you do 'bmod -n 1', the number of 'Processors Requested' is reduced and the job seems to fire off. Changing the 'Requested Resources' with 'bmod -R ...' does not seem to have an effect as long as 'Processors Requested' is greater than 1.

I am testing on a VM that thinks it has 24 cpus. Not sure if this problem is VM specific or just related to openlava...

— Reply to this email directly or view it on GitHub.

malachig commented 11 years ago

It seem that this has been working smoothly for a while now. Ref-align and other jobs in the standalone GMS no longer hang in a Pending state now that the 'Processors Requested' value is not set when submitting resource requests to openlava.

I believe the reference to openlava specific code that Scott makes above is here: Genome/InstrumentData/AlignmentResult/Bwa.pm

gatoravi commented 11 years ago

OK, i think the commit that Scott( or Obi?) made is this one, genome/tgi-workflow@dbd055c5ccb8f0c6ab04e57b794eeb114591f0f6

Looks like the ncpus option is still used but -n is disabled for OPENLAVA.