There is no collector load balance and fail over before 6.0.
First we have to configure collector group.
Let me explain how collector fail over works.
1: A Collector Group is shown – which includes both Collectors
2: The List of all Collectors in that Group is sent to all Collectors in the group from the Server
3: All Agents are configured to a single IP (a single Collector in the Collector Group)
4: The boostrap agent connects to the single Collector
5: The ‘real’ agent is downloaded to the Agents
6: The bootrap agent disconnects from the Collector
7: The ‘real’ agent connects to the single configured Collector
8: The Collector list is pushed to the Agents and some Agents are load balanced to the other Collector
9: The Collectors that are now load balanced to a different Collector connect to that Collector
10: Uninstrumened Byte/IL Code is sent to the Collectors and instrumented Byte/IL Code is sent back to the Agents
11: Meta data is sent from the Collectors to the Agents
12: Collector 1 goes down
13: Agents disconnect
14: Agents move to next Collector in list
15: Agents connect to new Collector
16: Agents send their Meta data to the new Collector
17: Agents send data to Server
18: Server requests new Meta data from Collectors
19: Collector sends Meta data to Server
That's how collector fail over works