[EMAIL PROTECTED] wrote on Wed, 13 Sep 2006 09:12 -0500: > Woah, I definitely didnt get things sent out, sorry about that Pete. > http://www.scl.ameslab.gov/~kschoche/netpipe3-pvfs.tar > > It's a very crude code for right now, and there are some obvious issues > with it, but it causes the hangs quite reproducibly. > There may be a hardcode that I thought I had patched, and not ever > exactly patched in the Makefile for the pvfs2 headers. > > Give it a try, it should break something, thats for sure.
I had to futz around a bit to get it to compile against recent PVFS head. Here's the tarball in case you want to check I did not break anything: http://www.osc.edu/~pw/netpipe3-pvfs-pw.tbz It's been running for 24 hours now, like this: while :; do date ; ./NPpvfs -d /pvfs-ib/foo ;\ ./NPpvfs -D /pvfs-ib/foo ; pvfs2-rm /pvfs-ib/foo ; done on one client, against six IO servers and one MD server. Sort of like your setup, but only 4X openib everywhere. Let me know if I should be using different arguments to provoke some bad behavior. There is one change in my tree that does not exist in the CVS yet that I can't definitely say would fix anything, but maybe it does. The change fixes some 100% load problems on the server. I'm starting a run now without that fix just too see if I can cause things to break like you can. Any advice on how to cause more damage? -- Pete _______________________________________________ Pvfs2-developers mailing list [email protected] http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
