Hi
Because ROMIO was passing uninitialized variables for PVFS2
distribution parameters (attribs.nr_dfiles and simple_stripe's
strip_size),  ROMIO was attempting to create a file striped across
324523452345 servers with a strip size of 3234532453345 bytes (or
something like that).  

Ok, so I fixed ROMIO to initialize those variables and the problem
goes away.  Sam and i were talking a bit, and we didn't really know
what "bad" values would be for strip size and nr_dfiles would be.  

We allow more datafiles than servers, but is there a point where that
number gets ridiculous?  100 dfiles per server? 1000?

How about strip size?  1 GB?  10 GB? 

I can sort of see a corner case where one might ask for extreme values
for nr_datafiles and strip size and I don't want to impose arbitrary
limits.  On the other hand, it would have been helpful if something in
PVFS2 had asserted or even logged the odd parameters for the create
request.  We clearly *do* have limits: PINT_distribute started
returning negative values, confusing PINT_process_request to no end :>

==rob

-- 
Rob Latham
Mathematics and Computer Science Division    A215 0178 EA2D B059 8CDF
Argonne National Labs, IL USA                B29D F333 664A 4280 315B
_______________________________________________
Pvfs2-developers mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers

Reply via email to