High Availability with 2 server nodes behind F5 - Health check for polling Tentacle

Legacy Scale
1

Server/Tentacle both v3.7.2. What were observed

  1. If only 1 server node is active behind F5 load balancer, health check on all polling Tentacles succeeds.
  2. If 2 server nodes are active behind F5, health check succeeds on some polling Tentacles (around half) and fails on others (i.e. the other half).

It appears that health check job is initiated on one server node, and Tentacles’ polling lands on the other server node (because of load balancer distributing the load), then first server node will think that the polling Tentacle never responded and therefore marks that Tentacle as unhealthy.

Can you confirm this could be the behavior and offer a solution?

2

Hi,

I’m sorry to hear you’re experiencing this issue. That shouldn’t be the case.

Would you be able to supply the Tentacle.Communication.TrustedOctopusServers section of your Tentacle.config file as documented here?

Regards,
Michael

3

Hi Michael - I didn’t know about this configuration on the polling Tentacle side. What I have is a single entry for the load balancer itself, instead of one entry for each of the server. I’ll modify the configuration as documented, then report back the result. I think it should resolve the issue.

4

We suspected that may be the problem.

It is unfortunate you have to explicitly enter each Octopus Server. We are currently investigating ways to make this more intuitive.

5

With the modification to explicitly adding one entry for each server, I’ve confirmed that the issue is resolved.

Look forward to a more intuitive solution.