[Bioclusters] NCBI BLAST & Sun GridEngine

Chris Dagdigian dag at sonsorol.org
Thu Jul 13 06:55:49 EDT 2006


This is a trivial method that will have its own problems but it will  
get you started:

Write a wrapper script for the NCBI program, one that knows how to  
talk to Grid Engine. Replace the real NCBI binary with the wrapper  
script so that the CGI program calls the wrapper instead of the real  
binary. The end result is your CGI's will call the grid aware wrapper.

The wrapper can be as simple as something like this:

  $!/bin/sh
  #
  # Grid Engine qrsh wrapper for blastall
  #
  # Run blastall via SGE 'qrsh' while preserving the command line  
arguments
  qrsh /path/to/real/blastall $*


That wrapper script will use SGE 'qrsh' to run the real blastall  
command while passing along all of the NCBI blastall arguments,  
database selections etc. that the user wanted to make use of. Qrsh is  
a program that will basically run your command ASAP on the least  
loaded node in the cluster. There are problems with qrsh, especially  
when the cluster is full (it will fail or exit with an error you will  
have to trap) but this method will at least get you up and running.

Qrsh will also direct its results to the STDOUT stream which is  
probably what your CGIs are expecting anway


Things to watch out for:

1. the cgi is running as the webserver user, you need to make sure  
the critical SGE environment variables are present in the CGI  
environment or defined explicitly in your wrapper script. A good way  
to test is to make sure you are easily able to run command line jobs  
as the webserver user









On Jul 12, 2006, at 5:56 PM, Ashwin Kotwaliwale wrote:

> Hi there,
>
>
>
> I am trying to setup a blast server and following several advices I  
> installed Sun Grid Engine. The problem is that I want to run BLAST  
> as web service unlike via command line, the reason being software  
> which I have will communicate with BLAST only via cgi calls. This  
> all works OK but I can’t get the jobs to be run on all the machines  
> in the grid. How does one configure www NCBI BLAST for use on SGE?
>
>
>
> Any help will be highly appreciated!
>
>
>
> Thanks in advance!
>
>
>
> Ashwin
>
> _______________________________________________
> Bioclusters maillist  -  Bioclusters at bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters



More information about the Bioclusters mailing list