Forum Discussion

k20's avatar
k20
Icon for Nimbostratus rankNimbostratus
Nov 14, 2017

Active standby pool members

I'm currently using Priority Group Activation method to accomplish my first objective and it's been working great. However, I have more requirements than just active standby real servers. The situation gets complicated when my primary server fails and later comes back online. When it fails, there is no question about it, all requests will hit the standby server. Here's a tricky part when the primary comes back online. Because of the Priority Group Activation that I have above, when the primary comes back, all traffic would go back to the primary. This is NOT what I want. My new requirement is that when the primary comes back, I still want my traffic to hit the standby unless I want to manually force the standby offline. Some people in this forum have suggested to use Destination Address Affinity in combination with Priority Group Activation. Now before I go into details, I wanted to be clear on my objectives.

 

  1. Primary and standby, all traffic goes to primary
  2. Important: ssh traffic, I don't want my connections to be closed or reset when failover occurs
  3. When primary fails then later comes back, I want my traffic (new and existing) stays on the standby until I manually take the standby offline.
  4. When both nodes are online (important, especially after the primary is recovered from a failure), I don't want requests to hit both servers due to Priority Group Activation and destination Address Affinity combination.

How do I achieve all of those objectives above? Please advise.

 

5 Replies

  • Enable manual resume on your health monitor.

     

    And change (under advanced configuration) your pool response to load balancing fail to reselect.

     

    Cheers,

     

    Kees

     

  • The solution is iCall

     

    You can use alertd to create a trigger when a pool member in your pool gets marked down.

     

    This triggers a tcl script that swaps the priority group activation settings for the pool members.

     

    iCall has a few documents and examples, so have a look.

     

  • Its not an answer for your question rather a question again.

     

    My Requirements, 1. Primary and standby, all traffic goes to primary. 2. When primary fails, I want my existing connections also switch to standby. 2. When primary fails then later comes back, I want my traffic (new and existing) go back to primary. 3. When the both nodes are down, I want all the existing and new connections should drop.

     

    Your configuration is almost matching for my requirement, could you share, which configuration should I have. Thanks in advance.

     

  • you can get your requirements just using:

     

    • Priority Groups to create an Active-Standby scenario.
    • Destination persistence/Universal persistence to not give traffic back to primary in case of failover.
    • Connection Mirroring (to avoid closing connections in case of failover)

    KR, Dario.