Frealign not going into reconstruction in SGE queue

Submitted by msaur on Mon, 2015-01-26 05:18

Forums

Frealign

Hi everybody,

I am currently installing Frealign on a new cluster and downloaded the NSF1 dataset in order to see if the program functions properly inside our queue. Would anyone perhaps have an idea as to why Frealign in an SGE queue would not go into the reconstruction when refining the whole dataset, but would go into the reconstruction when only using particles 1-100 (I haven't tried yet up to which number it would still work). It wouldn't work either if I give Frealign the full dataset with first_article 1 and last_particle 10002. It seems to me like some problem in the monitoring. Some time-out thing that works when everything is fast but does not work when the actual monitoring has to kick in over some time. Indeed I have errors inside the monitor_realign.log file:

tail: cannot open `/work/msaur/tests/nsf_frealign/scratch/130423' for reading: No such file or directory
tail: cannot open `130424' for reading: No such file or directory

I looked at the code in monitor_realign.com and cannot find the position at which the file $SCRATCH/${logfile} is actually written.

If anybody has experienced similar problems I would really appreciate some input. If more information about anything specific is needed I will do my best to provide it.

Best,

Michael

niko Mon, 2015-01-26 09:20

When installing the latest version of Frealign, please make sure that the mparameters file has the latest format. Frealign should be backward compatible but sometimes this might not work. To get the latest mparameters template, type

frealign_template

Then fill in all the values that are specific to your test or dataset.

It is possible that there is a timing issue that affects the monitoring of jobs. If this is the case, try killing the monitoring job (monitor_frealign.com) and see if the job completes.