I’m relatively new to this , and by a consultation of the costumer, where we will implement HA, I would be interested to know about the connectivity between the “Gad” and “interaction_servers” or “TServers”.
Someone may tell me something? ???
I search in the KB, but not found something that I clarify me the issue. :-[ :-[ :-[
I commented:
Given that we will implement “HA” in this client, I need to know:
1 - How is the connectivity between Gad and the interaction tservers or servers? Both the primary server as the Backup.
2 - If main Gad connection is lost, how it connects with the backup?
I have seen GAD load balancing implemented in a several different modes. GAD can utilize an Apache 2 configuration that distributes GAD traffic across several tomcat servlet engines; the use of a hardware load balancer, F5-BIG IP, and Windows Clustering.
Take a look in the 7.6 deployment guide for more detailed descriptions on pg. 162.
The Apache 2 configuration using the mod_jk connector is the quickest in my opinion.
GAD is not designed to be working in HA.
The problem comes from the Tomcat session ID which cannot be shared between several servers.
Here we have installed 3 instances of GAD on multiple servers, each connected to both primary and Backup IxnServer and TServer, and the customer can choose which one they want to connect to.
Advantage against load balancing, is that if they feel that there is something strange with one GAD (or slow !!!) they connect to the other one. ;D
Finally to lower the risk of issues, the GADs are not running the same exact version, in case one version has a bug the other has not.
1.GAD is not used the HA schema. Use the load balansing instead HA.
2. As and for another services - in general, you don’t need connect one services to primary and backup target services. only to primary.
3. So, coz no HA for GAD - if one of GAD instance fail, all agent connection is also broken down.
Loads of answers here that aren’t quite accurate. There are several layers being discussed and mixed up.
HA is possible through CIM. Agents always register with SIP Server or TServer, depending on what you use, IP or TDM and in turn register with the switch via either method. If you are planning to achieve HA for Genesys desktop, there are multiple ways in achieving this without extra H/W. We have GAD distributed across multiple data centers, also have all CIM components HA across the same data centers. Load balancing across centers; lose Site 1 GAD, then CIM redirects the lost sessions to Site 2.
Load balancing or clustering of your Web App Servers within a single data center is what many of you are speaking of here. Yes, there are several ways to manage clustering and load balancing. That discussion can fill several threads here for days. And there are plenty of great resources outside of Genesys that discuss load balancing with devices, Apache, etc.
The attached is a picture of what this looks like at a high level.