Opened 19 months ago

Closed 19 months ago

Last modified 19 months ago

#1486 closed enhancement (fixed)

Add an option to run the older bwa-mem instead of bwa-mem2

Reported by: Nicklas Nordborg Owned by: Nicklas Nordborg
Priority: major Milestone: Reggie v4.46.1
Component: net.sf.basedb.reggie Keywords:
Cc:

Description

There are some issues with bwa-mem2 that causes it to crash. It is some kind of memory-related error and the error message may differ from time to time. Typical are: double free or corruption, Segmentation fault and invalid pointer.

The original bwa-mem (version 0.7.17) has been tested on one of the problematic samples and it seems to be working.

Attachments (1)

bwa-mem2_crash.zip (18.7 KB ) - added by Nicklas Nordborg 19 months ago.
R1 and R2 with 250 read pairs that causes bwa-mem2 to fail

Download all attachments as: .zip

Change History (5)

by Nicklas Nordborg, 19 months ago

Attachment: bwa-mem2_crash.zip added

R1 and R2 with 250 read pairs that causes bwa-mem2 to fail

comment:1 by Nicklas Nordborg, 19 months ago

In 7178:

References #1486: Add an option to run the older bwa-mem instead of bwa-mem2

A new version of the container that include both bwa 0.7.17 and bwa-mem2 2.2.1

comment:2 by Nicklas Nordborg, 19 months ago

In 7179:

References #1486: Add an option to run the older bwa-mem instead of bwa-mem2

Added a configuration "parameter-set" that uses the original "bwa" aligner instead of "bwa-mem2". There are some minor changes in the script to make sure that "bwa-mem2" is selected by default and that log/progress messages uses the actual program name in the output.

comment:3 by Nicklas Nordborg, 19 months ago

Resolution: fixed
Status: newclosed

comment:4 by Nicklas Nordborg, 19 months ago

In 7180:

References #1486: Add an option to run the older bwa-mem instead of bwa-mem2

Updated the configuration with the new values for the number of cores to use for various analysis steps. Using 32 cores for bwa-mem2 turned out to be inconvenient since the computation nodes have 64 cores. By changing to 24 cores there is still room for a 16-core job or 2 8-core jobs to execute alongside the alignment on each node.

Note: See TracTickets for help on using tickets.