Issue after 2 node cluster tutorial

[Quietman]

I've setup a 2 node cluster following the tutorial posted, looking to generate a stable ha platform to replace our production FusionPBX (on Ubuntu).

The install and initial testing has gone well, but I've hit a wall when it's come to Call Center functionality.

My node servers are running 2 nic interfaces (1 x LAN- 10.15.20.0/24, and 1 x direct connected heartbeat - 10.100.100.0/31). I installed the platform using the up addresses From the heartbeat interfaces as the cluster ips, as I wanted to keep the bdr replication off the corporate LAN.

The FS domain is based on a virtual floating IP address, which sits on the Corp LAN. IP allocation is performed under a pacemaker/corosync setup.

All seemed well, but when I'm running operations on the Call Center app, the ha_monitor script is dumping ERR_NOT_FOUND into the fs log. Initially I thought it was an error with curl and a proxy setting, but examining the xml in the event console, shows the API event is calling the cluster IP as the Freeswitch-Domain tag. This is causing the API call to fail.

Also I have noticed that the Call Center creation process is not generating a BDR sync to the other cluster node.

This means, even if the API request was properly formatted, the secondary node would have no knowledge of the Call Center config I had created.

Anyone have any ideas on this, or can anyone point me to the right place in the config to investigate further, as the ha_monitor.lua, or xml_rpc config doesn't allow for setting the Freeswitch-Domain tag, or explain why the cluster node IP is being sent instead of the actual freeswitch domain.

Does the Call Center app run a separate db instance which isn't included in the bdr configuration?

Alternatively, can anyone else confirm if they have successfully run a ha config with a dedicated heartbeat, and FS domain based on a separate interface address.

Thanks in advance all.

 

[DigitalDaz]

Anything is possible but this is a simple two node cluster that is meant to be light, hence no replication of the freeswitch database. There is no attempt to use sofia recover etc to rescue calls. One server fails, all ongoing calls fail with the server. This means there is no need to use ha_monitor or anything like that. For a two node with floating IP I just call fs_cli -x "memcache flush" in the script that brings the floating IP up. That takes care of memcache.

 

[inform11]

Can I use the command fs_cli -x "sofia recover" in the script that brings the floating IP up ? Or is it better to use the command fs_cli -x "memcache flush" ?

 

[Luis Daniel Lucio Quiroz]

I've setup a 2 node cluster following the tutorial posted, looking to generate a stable ha platform to replace our production FusionPBX (on Ubuntu).

The install and initial testing has gone well, but I've hit a wall when it's come to Call Center functionality.

My node servers are running 2 nic interfaces (1 x LAN- 10.15.20.0/24, and 1 x direct connected heartbeat - 10.100.100.0/31). I installed the platform using the up addresses From the heartbeat interfaces as the cluster ips, as I wanted to keep the bdr replication off the corporate LAN.

The FS domain is based on a virtual floating IP address, which sits on the Corp LAN. IP allocation is performed under a pacemaker/corosync setup.

All seemed well, but when I'm running operations on the Call Center app, the ha_monitor script is dumping ERR_NOT_FOUND into the fs log. Initially I thought it was an error with curl and a proxy setting, but examining the xml in the event console, shows the API event is calling the cluster IP as the Freeswitch-Domain tag. This is causing the API call to fail.

Also I have noticed that the Call Center creation process is not generating a BDR sync to the other cluster node.

This means, even if the API request was properly formatted, the secondary node would have no knowledge of the Call Center config I had created.

Anyone have any ideas on this, or can anyone point me to the right place in the config to investigate further, as the ha_monitor.lua, or xml_rpc config doesn't allow for setting the Freeswitch-Domain tag, or explain why the cluster node IP is being sent instead of the actual freeswitch domain.

Does the Call Center app run a separate db instance which isn't included in the bdr configuration?

Alternatively, can anyone else confirm if they have successfully run a ha config with a dedicated heartbeat, and FS domain based on a separate interface address.

Thanks in advance all.

You are looking for a proper HA system, http://inside-out.xyz/technology/how-to-configure-freeswitch-for-ha.html
If you need more, contact me privately

 

[kp123]

can do recover call to node2 when service freeswitch stop on node1?

 

[DigitalDaz]

Probably not, I don't think I'm replicating the freeswitch db. If you know what you are doing though, this can be achieved.