This discussion has been locked. The information referenced herein may be inaccurate due to age, software updates, or external references.
You can no longer post new replies to this discussion. If you have a similar question you can start a new discussion in this forum.

Does it make sense to configure agents to a Load Balanced (LB) VIP that pushes to our APE?

Hello,

This is agent based communication.

We have about 870 agents that connect to various pollers. We wanted to ensure agent availability if poller it connects to goes down.

To me, it makes sense to create a LB VIP like the SNMP traps and just point the agents to a VIP and let the LB push to the pollers.

Does it make sense to configure agents to a LB VIP that pushes to our APE?

When an agent loses connection, I'm assuming the agent will try to reconnect every X amount of minutes to the VIP, thus  pointing to an available poller. 

Thanks!

  • From past experimentation with this concept there cannot be a LB between the pollers and agents.  The pollers themselves have to be expecting the agents, if they receive traffic from an agent they aren't assigned to in the DB they won't do anything with the data.  In the DB each agent can only be assigned to a single polling engine.

  • I assume we are only talking about Agent initiated communication.

    When you do the install, you are required to provide an Orion Polling engine server name, which could be the VIP, but I would guess that at some point in the communication, the agent looks up the setting assigned to the node for Polling Engine and then uses that address for future connections.  So that would be the name or IP address of the actual polling engine it's assigned to. 

    Now if that happens to be the name, then you could use the local hosts file to point it back to the VIP address and it might work.  But I would think they lookup the IP address in the Orion.Engines table.  Then communication would either just fail or route directly to the server depending on your F5 setup.

    We really need a agent developer to chime in, because everything I got is guesses.

    Assuming that all of that is even correct, then we could put in a feature request.  We could have a checkbox "Always use this address for agent connections", and it would always use the address you entered during the setup, which would be the VIP address.

    Or you could just change it to Server initiated connections and not need the VIP, but I understand that is not preferred in all situations.

  • Pretty sure the agents are locked down to a specific engine ID regardless of using server or agent initiated, but the best way to find out is to test it out.  I've seen many cases where there is a mismatch between what server the agent thinks it should send data to versus what poller thinks it should be getting agent traffic, and from wiresharks I've seen agent traffic show up on the incorrect server and that data never got added to the DB.

  • Thank you, this has answered my question!