qbatch

qbatch copied to clipboard

I'm thinking returncode of the job submission and job ids. This would depend on getting the properly parsable outputs.

gdevenyi

afterany dependency support

3

Currently `--depend` sets up dependencies so that all jobs must succeed before the submitted job runs (in PBS this is the `afterok` dependency), but sometimes you may want your final...

pipitone

Need alternative way to specify no memory request

1

Currently, memory request of 0 is used to indicate not making a memory request. SLURM reserves 0 to a special request of "give me all ram on the node". So...

gdevenyi

Implement --min-chunksize

2

If I have something like a joblist of 81 commands and my chunksize is 20, I get this unfortunate result of jobs with 40,40,1. This is bad for the cluster....

gdevenyi

When debugging cluster config its helpful to know exactly what inputs qbatch ended up with since a few different configs can result in similar looking job slicing.

gdevenyi

Re-implement command-line parameters and environment variables as ConfigArgParse interface

9

Right now we allow defaults via the environment. @andrewjankie makes a passing mention of config file defaults

gdevenyi

This is an unhelpfully imprecise ticket title as is, but let me explain here and we can perhaps break out separate tickets if that seems more useful. I think we...

pipitone

Figure out "live" testing

2

Have had a couple of bugs now that are only exposed by actual job submission. Need to figure out some kind of live testing using docker maybe?

gdevenyi

Need check for gnu-parallel vs moreutils parallel

1

Different code, same name. Need to parse a version string?

gdevenyi

Default chunk_size and cores to ppj

6

Back in issue #49 we discussed having chunk size defaulting to the ppj, but we lost track of that issue so I'm filing this. I'm thinking it the logic should...

pipitone