Checkpointing failures with BayesWave on LDG
bayeswave_post sometimes throws errors such as the following
gsl: interp.c:83: ERROR: x values must be monotonically increasing
Default GSL error handler invoked.
gsl: interp.c:38: ERROR: insufficient number of points for interpolation type
Default GSL error handler invoked.
This has been identified to be problem with the check-pointing step of bayeswave runs when the jobs are held/evicted during down-times etc. The files in the chains/
directory get under written or over written because of problems with I/O during the checkpointing phase.
The workaround to this, which is to manually check the chains/
files, remove spurious lines, and then rerun bwp is hit or miss. One suggested solution was to let the job run for 10 more minutes so that bayeswave can finish it's last I/O step.
Edited by Tyson Littenberg