[Gluster-users] Geo-replication causing input/output errors on nfs clients

[Sam Wouters]

Hi,

we have a setup with a 2 node gluster replication cluster where the 
volume is exported through the gluster build-in nfs service.
This volume is also replicated to a secondary site with geo-replication.

The problem is, that when clients are writing files larger than 10M, 
once every time the write gets aborted with a input/output error. At 
that moment, following lines appear in the gluster/bricks log:

[2013-07-01 12:41:39.800465] E [posix.c:2128:posix_writev] 
0-<vol>-posix: write  failed: offset 1069056, Bad file descriptor
[2013-07-01 12:41:39.800510] I [server3_1-fops.c:1414:server_writev_cbk] 
0-<vol>-server: 114834: WRITEV -2 (8e01b426-11dd-4147-9e5c-9256ca118be7) 
==> -1 (Bad file descriptor)

When geo-replication is stopped, these errors do not occur, so I suspect 
this is some kind of locking issue between the integrated nfs server en 
the crawler for the geo-rep.

Anybody any ideas how to circumvent this?

Version information of the gluster servers:

CentOS release 6.4 (Final)
glusterfs-3.3.1-1.el6.x86_64
glusterfs-server-3.3.1-1.el6.x86_64
glusterfs-fuse-3.3.1-1.el6.x86_64
glusterfs-geo-replication-3.3.1-1.el6.x86_64

regards,
Sam
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130704/8e1e3e3b/attachment.html>