You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It might be valuable to have a script to look for failed alignments and create a new input file based on those (unless you decide to change everything to one whole sample per computing node design). I did a run yesterday afternoon where many (but not all) of the tasks failed with error message like:
[gadi-cpu-clx-1277:315845:0:315991] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x2837de0)
[1626069937.336272947] [gadi-cpu-clx-0731.gadi.nci.org.au] [00:S] Task 4 was terminated by signal 11: ./align.sh Fastq_split/101.OSCC_30-P_HC5GHDSX2_S1_L001,OSCC_30-P,RamaciottiCFG,1a,illumina,HC5GHDSX2,1,8
Copying and pasting the command onto the login node and running bwa there revealed that the task above finished successfully and created a BAM file. Ben Menadue thought it was a case of "memory pressure" on the node at the time the anaysis ran. Today, submitting the same job with identical settings worked fine. I only changed the names of the output and error log files.
Given how low the address is which it attempted to access (0x2837de0 = 42171872 ~ 40MiB), my immediate suspicion is that an attempt to allocate memory has failed but your program hasn't checked for that. For example, malloc returns an address of NULL (i.e. 0) on failure, which can easily result in low-address segmentation faults if the program just assumes that the malloc succeeded. In such a situation, the bug in the program is that it didn't consider the case where the call to malloc failed, and just blindly used the return value even though it wasn't a valid address.
It seems that bwa can sometimes do some bad memory allocations, which the pipeline (nor bwa) guard against.
The text was updated successfully, but these errors were encountered:
It might be valuable to have a script to look for failed alignments and create a new input file based on those (unless you decide to change everything to one whole sample per computing node design). I did a run yesterday afternoon where many (but not all) of the tasks failed with error message like:
Copying and pasting the command onto the login node and running
bwa
there revealed that the task above finished successfully and created a BAM file. Ben Menadue thought it was a case of "memory pressure" on the node at the time the anaysis ran. Today, submitting the same job with identical settings worked fine. I only changed the names of the output and error log files.It seems that
bwa
can sometimes do some bad memory allocations, which the pipeline (norbwa
) guard against.The text was updated successfully, but these errors were encountered: