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.

APE HA Licences and APE HA design

Hi. Is this info still actual?

if you want to make an APE highly available, all you need is your original APE license used on your existing APE server, and one HA license that is used on the standby server.

We deploy MPE HA cluster on the main location (that's why we need 1 HA license) and we want to distribute APE HA (APE + APE failover) on different location.

I need second HA license for APE Cluster - am I right?
So total it would be: 1 HA lic. for MPE, 1 lic. for APE, 1 HA lic. for APE = 3 licences?

Also I'd like to know if APE HA design the same as MPE HA design? So you still need VIP for APE HA Cluster and all the info from https://documentation.solarwinds.com/en/Success_Center/orionplatform/Content/HA_Which_IP_is_the_Source.htm still actual for APE HA Cluster?

 

image.png

  • Yes that is all accurate. APE HA clusters work exactly the same as MPE HA clusters.

    You did leave off the actual licenses for your MPE modules such as NPM/SAM/etc though.

  • Hmm, got it, thank you.

    Is it possible to make MPE HA Cluster with only one APE? Or HA Cluster means 2xMPE + 2xAPE by default?

    In other words, can I attach lonely APE to my MPE HA pool?

    And one more related question: where can I find a detailed description of failover protection? (How VIP works, best practices for VM configuring, how heartbeat works etc.). I have read OrionPlatform_AdministratorGuide, but HA section is not very detailed.

  • One more thing from administrators guide. Single-subnet section:

    "5.Enter your VIP or virtual hostname. You can use both VIP and virtual hostname at the sametime or neither. If you use a virtual hostname, do not include the domain name. The VIP must be unassigned and on the same subnet as the primary and secondary servers."

    What does it mean - "The VIP must be unassigned"?

  • Two cases: normal situation (left) and failover (right)

    omavel_0-1583486086690.png

    As I understood, I must assign 2 ip's on NIC1 and 2 ip's on NIC2, and 1 of these 2 ip's is shared "VIP". But in that cases there will obvious be IP-collision. So I think I've missed something. 

    Also, solarwinds uses heartbeat or not?

  • I think the term pool implies flexibility to the configuration that doesn't actually exist. It is always a pair. MPE+MPEHABackup, APE+APEHABackup. There are no other combinations allowed, a server cannot perform multiple roles or anything else..
  • Unassigned, as in cannot be in use anywhere. YOU don't manually assign the VIP on a nic or anything like that. You run through the wizard and it does everything automatically.

    The heartbeat happens inside the database, there is a column that is being updated regularly and if it gets delayed by more than 30 seconds it assumed the server cannot reach the database and is effectively offline. The HA service that runs on each server will add or remove the VIP depending on who is the active HA server.

    Honestly its pretty simple once you see it in action, easiest way to figure it out is just to stand up two blank windows VM's and install the orion eval on the first and the HA eval on the second. The process is ridiculously simple and requires no special VM config ahead of time, just have an extra IP address picked out in your head and start the wizard.
  • Thank you for the answer.

    Also I have found this: "The SolarWinds HA software monitors the health of both servers in the pool, and both servers keep open communication channels over TCP port 5671 to exchange information. When a critical service goes down, such as the SolarWinds Information Service, the software starts the service. If the service goes down a second time within an hour, the software initiates a failover to the standby server."

    But what if heartbeat is OK (both servers can write to DB), but connection between them (TCP 5671) is lost? What is expected behavior of HA Cluster in this case?