Blog

Posted 2006/04/07

Binary install of R-GMA

As you may remember not even being able to install R-GMA was what kicked of the original javaGMA program. Well since I need to have R-GMA in order to do final benchmarks I have tried to stack the cards in my favour. Basically don’t build from source, don’t go against the grain and ask the developers about every detail.

I installed the operating system that the R-GMA folks use to develop the software (Scientific Linux 3.0.5) and then installed a binary release of R-GMA using apt-get. During this process I exchanged about a dozen emails with the R-GMA mailing list in order to work out what exactly I had to do. The process of working out what to try and how I got it all working is described in detail on the Wiki.

I can now run the R-GMA test scripts and pass everything.

This evening I am writing a hello world R-GMA client. Once I have that working I will try and extract the install of R-GMA and get it running on a normal Debian compute node. I want to run it on the same setup we have used for every other test and while the hardware would be the same Scientific Linux uses a heavily patched kernel and different versions of some core libraries.

I am sure it would have been possible to get it working on Debian directly (alien etc) but using the recommended system lets me get the R-GMA people to spell out what I need to do and if it all goes tits up going the recommended route makes it easier to report problems.