[maker-devel] Maker jobs hanging
Carson Holt
carsonhh at gmail.com
Fri Feb 21 11:56:50 MST 2014
Use 2.31. It has been tested to work without issue on several thousand
cpus. Also use OpenMPI for any jobs greater than 100 cpus. In addition,
OpenMPI can freeze on some systems without the following flag when using
perl based MPI programs --> -mca btl ^openib
Example --> mpiexec -mca btl ^openib -n 200 maker
Finally, never use MVAPICH2. It doesn't play well with perl, and freezes
whenever perl based MPI jobs extend across nodes (they run fine within a
single node though).
—Carson
On 2/21/14, 11:48 AM, "Greg Taylor" <gtaylor at bcgsc.ca> wrote:
>Hello,
> I'm having a problem with Maker_2.28 jobs hanging. I am annotating a 3Gb
>genome with predictors SNAP and Genemark, and using ABySS assembled
>RNA-seq data. To do this I am using 480 processors on our local cluster.
>Once a run begins, 479 contigs are started, as noted in the
>*_master_datastore_index.log file, the standard error log for the whole
>job looks normal, as do the run.log and run.log.child.0 for the daughter
>processes. This seems to be sequence dependent, as re-running contigs
>that hang doesn't help, the same contigs will always hang. I'm still
>looking into this myself, but it seems most if not all the jobs are stuck
>at the Blastx stage. If you have any suggestions, your help would be
>greatly appreciated.
>
>sincerely,
>Greg Taylor
>_______________________________________________
>maker-devel mailing list
>maker-devel at box290.bluehost.com
>http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
More information about the maker-devel
mailing list