Time limit in a supercomputer

[metxeberria-aranzadi]

Hi,

I’m running HybPiper on a Supercomputer, and I’m wondering what happens if the job reaches the time limit before it completes. Does it terminate automatically, and if so, is there a way to restart or resume the jobs efficiently?

I’d appreciate any guidance on the best approach to restarting or resubmitting the jobs while minimizing data loss.

Thanks in advance for your help!

Best,

[mossmatters]

Hi,

HybPiper can be restarted from one of four stages using the --start_from flag in hybpiper assemble :

[--start_from {map_reads,distribute_reads,assemble_reads,extract_contigs}]

To determine where HybPiper stopped, check the log file inside the HybPiper output directory. You should also use force_overwrite to ensure any files that were open when the HPCC killed the job are overwritten. I would recommend starting with the stage it failed on - for example if the run failed during the Spades assembly stage (most likely, since it's the step that takes the longest), you should re-run that sample with:

hybpiper assemble [.... other parameters here...] --start_from assemble_reads --force_overwrite

If the log file does not give a clear answer, you can also get a clue from the intermediate output files in the output directory. If you find the following:

spades_genelist.txt means that the map reads step completed
target_tallies.txt means the distribute reads step completed
exonerate_genelist.txt means the assemble_reads step completed
genes_with_seqs.txt means the whole pipeline completed

Therefore if exonerate_genelist.txt does not exist but target_tallies.txt and spades_genelist.txt are there, it likely means HybPiper failed during the extract contigs step and should be restarted from there.

Hope that helps!

[metxeberria-aranzadi]

Thank you, very much! It is very helpful!