[Pgcluster-general] Trouble failing over (2 nodes conf)

[Crespo, Daniel @ SDS]

Hi everybody,
 
I'm having troubles with failover.
Here is my configuration for 2 servers (nodes)
 
Node 1:
- DB Cluster_1
- Rep_1
 
Node 2:
- DB Cluster_2
- Rep_2
- Lb_2
 
My order of starting the processes is:
1) Node 1 and 2 up
2) Replicators 1 and 2 up
3) Load balancer up
 
After starting them up in the above order, everything works fine.
 
I tried putting the load balancer in the node 1, and everything worked
fine too. However, there is a difference:
- When putting the load balancer in node 1, and disconnecting the node 2
from the network, the load balancer keeps connection; BUT
- When putting the load balancer in node 2, and disconnecting the node 1
from the network, the load balancer doesn't keep the connection. I would
have to restart postgresql and pglb in order for the load balancer to
connect to its local dabase (DB Cluster_2).
 
My first impression is that the load balancer seems to connect to the
first database cluster written in the cluster.conf file. Is this
correct? If so, why is it called "load balancer"? It is not balancing
any load...
 
Any clues on how to failover successfully?
 
Daniel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://pgfoundry.org/pipermail/pgcluster-general/attachments/20070627/6c4c475e/attachment.html