Custom toolshed tool: job memory requirements

Question: Custom toolshed tool: job memory requirements

4.0 years ago by

Germany

Is it possible to add job requirements to (toolshed) tools?

For example, if a tool minimally requires >4gb of ram, ("java -Xmx4g") this will most of the time run fine on a non-cluster enabled galaxy instance. However, when using a cluster this might fail with "PBS: job killed: mem job total 4131052 kb exceeded limit 1048576 kb" due to the default job being specified for a low memory machine.

I understand a galaxy admin is able to add this to his job_conf.xml, and thus have absolute control, but when using the toolshed one might expect the tool just to work without fiddling with any xml configuration files.

toolshed • 1.1k views

ADD COMMENT • link •

modified 4.0 years ago by jmchilton ♦ 1.1k • written 4.0 years ago by Jelle Scholtalbers • 360

4.0 years ago by

Bjoern Gruening ♦ 5.1k

Germany

Bjoern Gruening ♦ 5.1k wrote:

Hi Jelle,

you can try to see if "_JAVA_OPTIONS" is set during execution of your tool. If it's not set you can provide your default settings. Try to document your tool in a way that gives a hint to the Galaxy admin that indicates that he should use "_JAVA_OPTIONS" to override your settings.

Hope this works for you,

Bjoern

ADD COMMENT • link written 4.0 years ago by Bjoern Gruening ♦ 5.1k

ow, this might indeed be the solution for the error stated... However, I'm actually more referring to an error similar to "PBS: job killed: mem job total 4131052 kb exceeded limit 1048576 kb". (sorry for the confusion). I updated the question.

ADD REPLY • link written 4.0 years ago by Jelle Scholtalbers • 360

4.0 years ago by

jmchilton ♦ 1.1k

United States

jmchilton ♦ 1.1k wrote:

It is not possible at this time - I have updated this (https://trello.com/c/HGEYPQf6) Trello card with that suggestion. In general deployers state fixed things about how jobs run that are not informed by the tool - I think finding the right way to templatize what the tool author suggest when the deployer is specifying things such as cluster parameters will be a real challenge - but it would definitely make these things more usable.

ADD COMMENT • link written 4.0 years ago by jmchilton ♦ 1.1k

Please log in to add an answer.

Similar posts • Search »