Running ChromA on SGE: Redis failed to start

[angelosarmen]

I tried to run consensus ChromA on an SGE cluster using 16 slots with 16 GB memory per slot (total 256 GB), but the job aborted after this warning:

2020-07-24 01:22:59,368 INFO resource_spec.py:204 -- Starting Ray with 154.88 GiB memory available for workers and up to 139.7 GiB for objects. You can adjust these settings with ray.init(memory=<bytes>, object_store_memory=<bytes>).
2020-07-24 01:22:59,620 WARNING services.py:923 -- Redis failed to start, retrying now.

I don't know if this has anything to do with ChromA using more processors than the available slots, as seen in sample.log:

2020-07-24 01:22:58,783:  Running Chrom in Info Mode
2020-07-24 01:22:58,784:  Validating Inputs
2020-07-24 01:22:59,332:  Inputs Validated
2020-07-24 01:22:59,334:  Processing File:1.bam
2020-07-24 01:22:59,334:  Processing File:2.bam
2020-07-24 01:22:59,334:  Processing File:3.bam
2020-07-24 01:22:59,334:  Processing File:4.bam
2020-07-24 01:22:59,334:  Processing File:5.bam
2020-07-24 01:22:59,334:  Processing File:6.bam
2020-07-24 01:22:59,334:  Processing File:7.bam
2020-07-24 01:22:59,335:  Running with 2 States: r=3, r=2
2020-07-24 01:22:59,337:  Running with 24 processors. Size of Plasma Storage 150000000000
2020-07-24 01:23:00,899:  Training on Regions

Does ChromA require exclusive use of the node as in the SLURM example? Looking at the code, this seems to be the case, as ray.init() is called with the number of processors - 1.

PS I tried asking the cluster for 24 slots but I would have to wait.

[marianogabitto]

Dear Angelos, yes, I coded ChromA thinking that it will use just 1 exclusive node with many processors.

     I am not familiar with the SGE cluster but I bet that it uses qsub to submit jobs. Can you try the option 

PBS –l nodes=1:ppn=24

     and let me know if it works? 

     Also, can you try running just one file to see if installation and everything are all right? 

      Please, my email is mariano.gabitto@gmail.com . Write to me as I am interested in seeing if this works in SGE . 

Thanks,

Mariano