[Beowulf] glusterfs and openmpi/mpich problems

[Joe Landman]

Gerry Creager wrote:
> We've been working with gluster of late, on our high throughput cluster 
> (126 nodes, gigabit connected).  We did some tweaking recently, and now, 
> my test code, an instance of WRF on 128 cores, just sorta dies.
> 
> More specifically, it takes 19 minutes to write the first 403MB file to 
> disk, while various tasks are mindlessly using CPU time, but only the 
> initial output file appears to get written.
> 
> Does anyone have any history with gluster who might be willing to offer 
> some help/hints?

Yes.  What translators are you using?  What is your back-end file system 
underneath glusterfs?  Which kernels are you using?



-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web  : http://www.scalableinformatics.com
        http://jackrabbit.scalableinformatics.com
phone: +1 734 786 8423 x121
fax  : +1 866 888 3112
cell : +1 734 612 4615