[Gluster-users] Server Side AFR and RoundRobin DNS

[Benjamin Long]

Greetings All,

	I've been fighting with a very strange problem after following the 
instructions at:
http://www.gluster.com/community/documentation/index.php/High-
availability_storage_using_server-side_AFR

I'm using Gluster 2.0.7 on Debian Lenny servers, and Kubuntu Karmic guest. 
Here are the symptoms:
Start Server1
Start Server2
Mount on client at /export
  * Server1 says:
[server-protocol.c:7065:mop_setvolume] server: accepted client from 
10.10.3.33:1023
  * Server2 says:
[server-protocol.c:7065:mop_setvolume] server: accepted client from 
10.10.3.33:1021

cd to /export
run 'ls'
  * Server 1 says:
[2009-10-05 17:54:54] E [server-protocol.c:5020:server_releasedir] server: fd 
- 0: unresolved fd
  * Server 2 says nothing.

run 'touch testfile'
  errors with:
touch: closing `testfile': Invalid argument
  * Client log says:
W [fuse-bridge.c:882:fuse_err_cbk] glusterfs-fuse: 87: FLUSH() ERR => -1 
(Invalid argument)
W [fuse-bridge.c:882:fuse_err_cbk] glusterfs-fuse: 89: FLUSH() ERR => -1 
(Invalid argument)
  * Server 1 says:
E [server-protocol.c:4119:server_flush] afr: invalid argument: state->fd
E [server-protocol.c:4119:server_flush] afr: invalid argument: state->fd
  * Server 2 says nothing


Now, the strange thing is, I can restart EITHER server and this problem goes 
away. Note that is a restart, not a stop. The server comes back up and 
replicates as I expect it to. I do not have this issue with client side 
replication, but it's much slower.

I've posted my complete config and more log details at:
http://pastebin.ca/1595892

-- 
Benjamin Long