Cluster Install - Trouble Finding Drmaa Egg

Question: Cluster Install - Trouble Finding Drmaa Egg

6.1 years ago by

mailing list • 840 wrote:

I'm trying to install Galaxy on our SGE cluster (using the Unified method). I installed galaxy-dist, and now I'm on this section: drmaa egg The drmaa egg is provided by Galaxy, but you must tell it where your resource manager's DRMAA library is located, and this is done with the $DRMAA_LIBRARY_PATH environment variable: % export DRMAA_LIBRARY_PATH=/galaxy/lsf/7.0/linux2.6-glibc2.3-x86_64/l ib/libdrmaa.so.1.0.3 % export DRMAA_LIBRARY_PATH=/galaxy/sge/lib/lx24-amd64/libdrmaa.so.1.0 However I can't locate the the file libdrmaa.so.1.0.3 anywhere. In fact I don't have have an lsf nor sge directory within galaxy-dist. Thanks, Greg

galaxy • 2.5k views

ADD COMMENT • link •

modified 6.1 years ago by Scott McManus • 130 • written 6.1 years ago by mailing list • 840

6.1 years ago by

Scott McManus • 130

Scott McManus • 130 wrote:

Have you installed SGE on the same machine? If so, then you should be able to do a "locate libdrmaa.so", which should point to SGE's DRMAA library. For example, on a machine that I use, I see this: $ locate libdrmaa.so /usr/lib/libdrmaa.so /usr/lib/libdrmaa.so.1.0 $ ls /usr/lib/libdrmaa* /usr/lib/libdrmaa.so /usr/lib/libdrmaa.so.1.0 $ ls -la /usr/lib/libdrmaa.so lrwxrwxrwx 1 root root 15 Apr 18 2012 /usr/lib/libdrmaa.so -> libdrmaa.so.1.0 You may also want to try "updatedb" (again, assuming you're using Linux) to index the files on your machine. Finally, if you can't find it, then you should check if you actually have SGE installed on that machine. I like to use "dpkg" on Ubuntu/Debian and "yum" on Centos to manipulate packages. -Scott

ADD COMMENT • link written 6.1 years ago by Scott McManus • 130

I forgot to add my value of DRMAA_LIBRARY_PATH. Note that I just need DRMAA_LIBRARY_PATH to point to the library being used. So in the example below, I had this: /usr/lib/libdrmaa.so -Scott

ADD REPLY • link written 6.1 years ago by Scott McManus • 130

Thanks Scott. I did turn up: /sge/8.0.1p4/lib/lx- amd64/libdrmaa.so.1.0 I guess that would be it. For a followup question ,how do I make the DRMAA_LIBRARY_PATH variable permanent on the machine? -Greg

ADD REPLY • link written 6.1 years ago by mailing list • 840

And am I correct in the thinking that only the machine hosting the galaxy web interface and submitting jobs needs the export DRMAA_LIBRARY_PATH= variable? The normal nodes running jobs don't need this, right? Thanks, -Greg

ADD REPLY • link written 6.1 years ago by mailing list • 840

Hi Greg, That's correct. If you're starting Galaxy with an init script, I'd put the export there. --nate

ADD REPLY • link written 6.1 years ago by Nate Coraor ♦ 3.2k

Well, I want it to ultimately run under Apache. Does it still make sense to go in an init script? thanks again, Greg

ADD REPLY • link written 6.1 years ago by mailing list • 840

Yes, even when running behind Apache, you'll need an init script (or similar method) to start the Galaxy server. --nate

ADD REPLY • link written 6.1 years ago by Nate Coraor ♦ 3.2k

Ok, that makes sense. Would you mind sharing how you set up your init script? What goes in it, where you place it? Thanks again, Greg

ADD REPLY • link written 6.1 years ago by mailing list • 840

There are some sample init scripts in the contrib/ directory of the distribution. You should be able to place the export near the top, where other variables are set. --nate

ADD REPLY • link written 6.1 years ago by Nate Coraor ♦ 3.2k

You can just start a screen session on the head node, cd into the galaxy directory and run "sh run.sh --reload" to start the processes or write a script to handle Galaxy in the daemon mode at any degree of granularity. For instance, here is my shell script that I use to start/stop/restart manager/runner, web runners or everything together as needed: http://pastie.org/5191826 Regards, Alex

ADD REPLY • link written 6.1 years ago by Oleksandr Moskalenko • 30

Similar posts • Search »