Gluster users - Jul 2015 - Mount Fails when 1 of 2 Replicas is Down (GlusterFS 3.7.2)

I am able to reproduce a problem, which I think may be a bug, where if 1 of
the 2 replica servers for a volume is down, clients are unable to mount the
volume. I notice that if the replica that is down is on the same subnet as
the client, the client fails to mount the volume, but if the replica that
is down is on a different subnet, the client fails over properly and mounts
the volume.

Here are the errors from the server that is still up when the client is
unable to mount the volume when the replica on the same subnet as the
client is down. Ideas? Should I open a bug?

[2015-07-01 05:43:08.428657] W [socket.c:923:__socket_keepalive] 0-socket:
failed to set TCP_USER_TIMEOUT -1000 on socket 21, Invalid argument
[2015-07-01 05:43:08.428710] E [socket.c:3015:socket_connect] 0-management:
Failed to set keep-alive: Invalid argument
[2015-07-01 05:43:08.429260] E [socket.c:3071:socket_connect] 0-management:
connection attempt on 10.1.0.100:24007 failed, (Connection refused)
[2015-07-01 05:43:08.429362] W [socket.c:642:__socket_rwv] 0-management:
writev on 10.1.0.100:24007 failed (Success)
[2015-07-01 05:43:08.429623] E [rpc-clnt.c:362:saved_frames_unwind] (-->
/usr/lib64/libglusterfs.so.0(_gf_log_callingfn+0x199)[0x7fd3d3470d59] (-->
/usr/lib64/libgfrpc.so.0(saved_frames_unwind+0x1ae)[0x7fd3d323f43e] (-->
/usr/lib64/libgfrpc.so.0(saved_frames_destroy+0xe)[0x7fd3d323f53e] (-->
/usr/lib64/libgfrpc.so.0(rpc_clnt_connection_cleanup+0xab)[0x7fd3d3240e6b]
(--> /usr/lib64/libgfrpc.so.0(rpc_clnt_notify+0x168)[0x7fd3d3241418] )))))
0-management: forced unwinding frame type(GLUSTERD-DUMP) op(DUMP(1)) called
at 2015-07-01 05:43:08.429387 (xid=0x5f)
[2015-07-01 05:43:08.429654] E
[glusterd-handshake.c:2001:__glusterd_peer_dump_version_cbk] 0-: Error
through RPC layer, retry again later
[2015-07-01 05:43:08.429820] E [rpc-clnt.c:362:saved_frames_unwind] (-->
/usr/lib64/libglusterfs.so.0(_gf_log_callingfn+0x199)[0x7fd3d3470d59] (-->
/usr/lib64/libgfrpc.so.0(saved_frames_unwind+0x1ae)[0x7fd3d323f43e] (-->
/usr/lib64/libgfrpc.so.0(saved_frames_destroy+0xe)[0x7fd3d323f53e] (-->
/usr/lib64/libgfrpc.so.0(rpc_clnt_connection_cleanup+0xab)[0x7fd3d3240e6b]
(--> /usr/lib64/libgfrpc.so.0(rpc_clnt_notify+0x168)[0x7fd3d3241418] )))))
0-management: forced unwinding frame type(GF-DUMP) op(NULL(2)) called at
2015-07-01 05:43:08.429395 (xid=0x60)
[2015-07-01 05:43:08.429845] W [rpc-clnt-ping.c:204:rpc_clnt_ping_cbk]
0-management: socket disconnected
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20150630/a3cfc016/attachment.html>

On Tuesday 30 June 2015 10:56 PM, Gabriel Kuri wrote:
> I am able to reproduce a problem, which I think may be a bug, where if 1
> of the 2 replica servers for a volume is down, clients are unable to
> mount the volume. I notice that if the replica that is down is on the
> same subnet as the client, the client fails to mount the volume, but if
> the replica that is down is on a different subnet, the client fails over
> properly and mounts the volume.
>
> Here are the errors from the server that is still up when the client is
> unable to mount the volume when the replica on the same subnet as the
> client is down. Ideas? Should I open a bug?
>
> [2015-07-01 05:43:08.428657] W [socket.c:923:__socket_keepalive]
> 0-socket: failed to set TCP_USER_TIMEOUT -1000 on socket 21, Invalid
> argument
> [2015-07-01 05:43:08.428710] E [socket.c:3015:socket_connect]
> 0-management: Failed to set keep-alive: Invalid argument
> [2015-07-01 05:43:08.429260] E [socket.c:3071:socket_connect]
> 0-management: connection attempt on 10.1.0.100:24007
> <http://10.1.0.100:24007> failed, (Connection refused)

This points to the client not being able to talk to glusterd on 
10.1.0.100. Is glusterd running on this node and if yes, can port 24007 
be reached from the client machine?

Regards,
Vijay