Significantly different runtimes for HaplotypeCaller on same BAMs & region

trgalltrgall Posts: 13Member

I have a region which worked fine, but now doesn't seem to work. Here is a part of the run using version 2.5
17:66743059-81195210.it1hc.stdout-INFO 21:59:30,131 ProgressMeter - 17:73500213
17:66743059-81195210.it1hc.stdout-INFO 22:00:30,949 ProgressMeter - 17:73500549
17:66743059-81195210.it1hc.stdout-INFO 22:01:31,026 ProgressMeter - 17:73500549
17:66743059-81195210.it1hc.stdout-INFO 22:02:31,045 ProgressMeter - 17:73500549
17:66743059-81195210.it1hc.stdout:INFO 22:03:31,063 ProgressMeter - 17:73500593
17:66743059-81195210.it1hc.stdout-INFO 22:04:31,098 ProgressMeter - 17:73501295

As you can see it made it by 17:73500549 in about 3 minutes. I am rerunning the same BAMs in the same region with version 2.6, and now it has been stuck on 17:73500549 for >12 hours. Even when rerunning the same version I have noticed that usually runs take about the same time, but every so often are orders of magnitude longer. I am using -dcov 200, so I know there is some sampling variation (although in this region everyone is between 50-100 reads coverage), but a difference of 3 minutes to > 600 minutes seems excessive.

Any suggestions on making runtimes for regions more predictable?



Best Answer


  • Geraldine_VdAuweraGeraldine_VdAuwera Posts: 9,278Administrator, Dev admin

    Hmm, that does seem excessive. Could you please try re-running that region with the latest nightly build? If you still see problematic runtimes, we'll ask for a test file that reproduces the problem, so we can debug it locally and find out what is causing the slowdown.

    Geraldine Van der Auwera, PhD

  • ebanksebanks Broad InstitutePosts: 691Member, Administrator, Broadie, Moderator, Dev admin

    How many samples/BAMs are you running with?

    Eric Banks, PhD -- Director, Data Sciences and Data Engineering, Broad Institute of Harvard and MIT

  • trgalltrgall Posts: 13Member

    ~80 samples with ~110 BAMs.

Sign In or Register to comment.