LSF10 incompatible and cannot submit jobs

[iiiir]

Hi,
We recently upgraded our LSF version from 8 to 10, and SJM no longer work.

There is no error message, but will have some "core dump" error as there will be a binary "core.123456" file been produced, and it much bigger than the original job file.
$ cat example4.sjm
job_begin
name jobA
time 1h
memory 500m
queue normal
project CompBio
cmd_begin
echo "hello from job jobA";
cmd_end
job_end

job_begin
name jobB
time 30m
memory 1g
queue normal
project CompBio
cmd_begin
echo "hello from job jobB";
cmd_end
job_end

order jobB after jobA

$ ~/app/SJM/src/sjm example4.sjm
Status file: example4.sjm.status
Log file: example4.sjm.status.log
Running jobs in the background....

$ ls -lrt
total 7200
...
-rw-r--r-- 1 xxxxx xxxxx 305 Mar 6 14:05 example4.sjm
-rw-r--r-- 1 xxxxx xxxxx 339 Mar 7 12:14 example4.sjm.status
-rw-r--r-- 1 xxxxx xxxxx 83 Mar 7 12:14 example4.sjm.status.log
-rw------- 1 xxxxx xxxxx 11247616 Mar 7 12:14 core.52031

We did some debug and it seems the job submitting has something wrong:
LS_LONG_INT jobId = lsb_submit(&req, &reply);
but still not sure what exactly went run.

I wonder if you can provide some insights?

Thanks a lot,

[nathankw]

Hi Shuoguo,

We don't have an LSF system to test on at present and our support for debugging SJM on systems other than SGE is very limited right now. We will likely be testing on SLURM later in the year as we transition our cluster to SLRUM. Theoretically SJM should run fine on DRMAA 1.0 compliant systems, nonetheless, as new versions come out for grid schedulers it may be that some of that backward compatibility is broken and SJM would need some patches to keep up with the flow.

The best I can suggest at present is to look at the SJM source code if you or a colleague has the time and if you find a patch that works for you on LSF you can create a pull request. Maybe a good place to start is the documentation for DRMAA support in the LSF version that you are using, and how that changed since the last version of LSF you were running.