Time limits on UnifiedGenotyper (UG)

MonkolMonkol BostonMember, Broadie

Hi

I'm currently having problems with UG not gracefully quitting and creating a valid truncated bcf/vcf file using the -maxRuntime and -maxRuntimeUnits options.

To test I have set -maxRuntime 1 -maxRuntimeUnits MINUTES and the process has run for a few days already!

I have two questions that may help resolve this problem:
1. Are actual real life minutes being measured or CPU minutes?

  1. How often is the runtime checked? (i.e. exceedsRuntimeLimit function called). As it appears that each of the jobs that are running over the time limit are stuck on a particular site and if it gets called after each site/interval is called then this may explain why it may be going over the time limit.

Thanks!

Answers

  • Mark_DePristoMark_DePristo Broad InstituteMember admin

    The max runtime is only enforced in the traversal, so you actually have to make some progress in the GATK before it will leave. It's something like this:

    initializeEngine()
    for ( each base ) 
      doSomeWork(base)
      if time limit exceeded exit
    

    That way you ensure you get good results (so you should never have truncated bcf/vcf files in the sense that they are malformed). But what it does mean is that if the GATK hangs in initializeEngine() or doSomeWork() you'll never leave. The Broad infrastructure is having major problems now -- you should upgrade to GATK 2.5 which has some protection against the infrastructure failures.

Sign In or Register to comment.