Strange behavior in HA mode

Hi!

I setup a ha cluster with master+slave following this tutorial: http://www.cubrid.org/manual/843/en/CUBRID%20HA.

I called my nodes : cubrid and cubrid2.

All nodes have this in cubrid.conf:

[@demodb]
ha_mode=on
force_remove_log_archives=no
log_max_archives=100

cubrid_ha.conf:

[common]
ha_copy_sync_mode=sync:sync
ha_port_id=59901
ha_node_list=demodb@cubrid:cubrid2
ha_db_list=demodb
ha_apply_max_mem_size=500

cubrid@cubrid:~$ cubrid heartbeat status
@ cubrid heartbeat list
 HA-Node Info (current cubrid, state master)
   Node cubrid2 (priority 2, state slave)
   Node cubrid (priority 1, state master)
 HA-Process Info (master 882, state master)
   Applylogdb demodb@localhost:/var/opt/cubrid/databases/demodb_cubrid2 (pid 3342, state registered)
   Copylogdb demodb@cubrid2:/var/opt/cubrid/databases/demodb_cubrid2 (pid 3340, state registered)
   Server demodb (pid 3123, state registered_and_active)


cubrid@cubrid2:~$ cubrid heartbeat status
@ cubrid heartbeat list

 HA-Node Info (current cubrid2, state slave)
   Node cubrid2 (priority 2, state slave)
   Node cubrid (priority 1, state master)
 HA-Process Info (master 2478, state slave)
   Applylogdb demodb@localhost:/var/opt/cubrid/databases/demodb_cubrid (pid 2704, state registered)
   Copylogdb demodb@cubrid:/var/opt/cubrid/databases/demodb_cubrid (pid 2702, state registered)
   Server demodb (pid 2485, state registered)

When I do some benchmark on the node "cubrid2", I can see node "cubrid" sending the data to cubrid2 then cubrid2 to my client. Using "top" command I can see node "cubrid" becomes busy with many INSERT, where as node "cubrid2" idle.

When I shutdown the node "cubrid", "cubrid2" becomes the new master but I can't send any query to cubrid2, it times out (trying to fetch data from node "cubrid"). When node "cubrid" is back, cubrid2 answer (I have to reload the heartbeat).

I'm using CUBRID version 8.4.3 from Ubuntu (12.04) repository

Thanks!