The goal of this document is to outline vendor agnostic data lifecycle when using a load balancer with HCP and the best practices for configuring the load balancer under specific requirements.
Using Active/Passive links, two pools are required, one for the primary HCP and one for its replica.
What must be done when the primary HCP fails? It should wait untill HCP is failed over (whether automatically or manually). Do the LB pools need renaming or re-population?
Kind regards, Henk
I’m not familiar with the functionality of the ZEN load balancer, and if it supports failover pool functionality like F5 and brocade do. In general one way to approach high availability is to use a failover pool on the load balancer. This allows the primary load balancer to re-direct traffic to the replica load balancer if the primary load balancer detects the primary HCP is offline. With this, no failover is required on the HCP as long as replicated namespaces are configured to accept requests redirected from other systems.
So no pools would need to be re-named or re-populated if you were using a failover pool.
Please see the section of the doc labeled: HCP High Availability with Load Balancers
HCP replication helps ensure that access to data stored on an HCP system is not disrupted if that system becomes unavailable. This high availability is achieved by means of a replication link between that system and another HCP system. If one of the two systems on a replication link becomes unavailable, client requests to that system can automatically be redirected to the other system on the link.
I have read this (v1.1, April 2018) document, it contains much detailed information. Still, the chapter High Availability with Load Balancers, instructs to "Set up a dedicated load balancer for each HCP system on an active/active link. My question is: how to setup (ZEN) load balancers for Active/Passive links. Manual intervention is not a problem, although automatic might be even better.
There would be no difference with and active active vs an active passive beyond the fact that you would not have seamless failover and would need to failover the HCP replication link.
Retrieving data ...