[Gluster-users] EBADFD with large number of concurrent files

[Brian Candler]

On Tue, Aug 07, 2012 at 05:18:24AM -0400, Shishir Gowda wrote:
> Can you please provide the client(mnt) log files?
> 
> Also if possible can you take the state dumps and attach them in the mail
> "gluster volume statedump <volname>"
> The o/p will be files in /tmp/<brick-path>.<pid>.dump.x

It is now looking like it may be a strange hardware failure on one of the
two servers; one of its NIC interfaces has dropped off (although we were
using the other one for traffic).  However, running the same job using a
non-distributed volume on the other node worked fine.

I will try to replicate this with the suspect node. If it's a hardware
problem then I apologise for the noise; if it's a software problem I'll do
the debug you suggest.

Thanks,

Brian