IOError: bad message length

[cwhelan]

I am having trouble getting NAIBR to complete on one of my samples. I see the following message in the logs:

Process PoolWorker-1:
Traceback (most recent call last):
  File "/broad/software/free/Linux/redhat_6_x86_64/pkgs/python_2.7.1-sqlite3-rtrees/lib/python2.7/multiprocessing/process.py", line 232, in _bootstrap
    self.run()
  File "/broad/software/free/Linux/redhat_6_x86_64/pkgs/python_2.7.1-sqlite3-rtrees/lib/python2.7/multiprocessing/process.py", line 88, in run
    self._target(*self._args, **self._kwargs)
  File "/broad/software/free/Linux/redhat_6_x86_64/pkgs/python_2.7.1-sqlite3-rtrees/lib/python2.7/multiprocessing/pool.py", line 73, in worker
    put((job, i, result))
  File "/broad/software/free/Linux/redhat_6_x86_64/pkgs/python_2.7.1-sqlite3-rtrees/lib/python2.7/multiprocessing/queues.py", line 366, in put
    return send(obj)
IOError: bad message length

, sometimes multiple times. After hitting one of these errors NAIBR continues running but eventually stops logging any progress updates and appears to get into a 'stuck' state in which it never terminates. Any thoughts on how to avoid this error?

My config file is below:

# minimum mapping quality (default=40)
min_mapq=40

# input bam file
bam_file=IN_DIR/outs/phased_possorted_bam.bam

# output directory (default=.)
outdir=OUT_DIR

# list of intervals not to be intcluded in analysis (default=None)
#blacklist=example/example_blacklist.bed

# list in BEDPE format of novel adjacencies to be scored by NAIBR (default=None)
#candidates=example/example_candidates.bedpe

# maximum distance between read-pairs in a linked-read (default=10000)
d=10000

# minimum size of structural variant (default=lmax)
min_sv=1000

# number of threads (default=1)
threads=4

# minimum number of barcode overlaps supporting a candidate NA (default = 3)
k=3