[Gluster-users] Peers keep disconnecting

[Paul Johnston]

Hi
I'm setting up a simple test rig, two bricks and a single controller.
All on machines running Fedora 16 with the latest version of gluster from the Fedora repos.
[root at h-s-h003839 ~]# glusterfs --version
glusterfs 3.2.5 built on Nov 23 2011 20:48:30

Looking at the peers:

gluster> peer status
Number of Peers: 2

Hostname: h-s-h003837
Uuid: 440b4101-bf78-4899-b118-ab2e3079bb5e
State: Peer in Cluster (Disconnected)

Hostname: h-s-h003839
Uuid: dbed62e1-6cd2-4114-95e1-6e9e461b8af8
State: Peer in Cluster (Disconnected)
gluster> 

Restarting glusterd on the peers occasionally gives a Connected state but this soon reverts to Disconnected:

gluster> peer status
Number of Peers: 2

Hostname: h-s-h003837
Uuid: 440b4101-bf78-4899-b118-ab2e3079bb5e
State: Peer in Cluster (Disconnected)

Hostname: h-s-h003839
Uuid: dbed62e1-6cd2-4114-95e1-6e9e461b8af8
State: Peer in Cluster (Connected)

I removed iptables from the machines and this does not help. All machines can resolve each other.
Below is a show selection of ping results
[root at h-s-h003838 ~]# ping -c 1 h-s-h003837
PING h-s-h003837 (130.88.105.63) 56(84) bytes of data.
64 bytes from h-s-h003837 (130.88.105.63): icmp_req=1 ttl=64 time=0.479 ms

--- h-s-h003837 ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 0.479/0.479/0.479/0.000 ms
[root at h-s-h003838 ~]# ping -c 1 h-s-h003839
PING h-s-h003839 (130.88.105.78) 56(84) bytes of data.
64 bytes from h-s-h003839 (130.88.105.78): icmp_req=1 ttl=64 time=0.231 ms

--- h-s-h003839 ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 0.231/0.231/0.231/0.000 ms
[root at h-s-h003838 ~]# 

Last March there was a similar problem raised by James Burnash under the title of "What does this error mean"


Anyone got any ideas of what might be causing this, more system information available on request.

Cheers Paul