Thanks for getting the log online. It looks like the pvfs2-client-core process is crashing for some reason, possibly hitting an assertion?

If you stop pvfs2-client and pvfs2-client-core and just start pvfs2-client-core by itself (making sure _not_ to use the --child option), then it should generate a core file if it crashes. You will have to manually restart pvfs2-client-core after that, but the core file might give us more information about what went wrong.

Is this only happening on your head/login node and not the compute nodes? I don't know of any problems with the du command, but there may have been something else that led up to the problem

-Phil

Phil Carns wrote:
Hi Jim,

Sorry to hear that you have having problems again. Could you check the url for your client log? Its not working for me right now.

thanks,
-Phil

Jim Kusznir wrote:
Hi all:

pvfs2 has been crashing on me a lot...its no longer taking down the
system anymore, but it is going completely unresponsive, and requiring
me to reboot the computer to get it to function again.

I managed to grab a pvfs2 client log, which I'm posting at:

http://www.eecs.wsu.edu/~kusznir/pvfs2-client-crash.log

As far as I can tell, the crash occurred with a user running du -sh on
some directory inside pvfs2.

This has been happening frequently.  It does not appear to be server
related, as all my compute nodes are still able to communicate with
pvfs2 while the head node is not.  Any attempt to do any operations
(including cd'ing) in pvfs2 simply hang indefinitely.

--Jim
_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users

_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users

_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users

Reply via email to