Add a scaling guide under the high availability topic

[shawkins]

Description

We don't provide guidance around scale up / down. It would be good to mention that infinispan must do rebalancing as members join / leave. It would also be good to describe in rough terms when scale up / down should be done in response to cpu, memory, or throughput limitations.

[shawkins]

@vmuzikar @keycloak/core @ryanemerson I'd like for there to be a topic inbetween the getting start guides and the HA guides around scaling / performance. It would cover at least this issue and #29391. It could be laid out roughly like:

Scaling and Performance

Intro and when is scaling needed...

Vertical scaling:

• link to ha concepts-memory-and-cpu-sizing
• link to ha concepts-threads
• link to ha concepts-database-connections
• could also mention the vertical pod autoscaler

Horizontal Scaling:

• cover how this also improves availability and link to the HA guide
• link to the operator as the prefered solution, could mention what issues it solve wrt discovery, etc.
• mention embedded infinispan single AZ consideration from Support pod affinity settings in the Keycloak Operator #29258
• mention node anti-affinity from Support pod affinity settings in the Keycloak Operator #29258 (or this could be strictly an HA topic)
• mention scale up / scale down considerations to satisfy this issue, and probably provide a horizontal pod autoscaler example. Could also mention that using external infinispan addresses the rebalancing concern.

Does something like this seem worthwhile and/or is it covered sufficiently by the HA guide?

[ryanemerson]

+1. These are all valid considerations for non-ha deployments, so I think it makes sense we document these topics outside of the HA guide.

• mention node anti-affinity from Support pod affinity settings in the Keycloak Operator #29258 (or this could be strictly an HA topic)

My understanding is that HA is related to multi-site deployments. I think anti-affinity is an important consideration for single site deployments, so I would include this in your proposed guides.

[kami619]

@shawkins thank you for logging this issue, I myself was contemplating on something like this to be added to our roadmap. We shall definitely bring this to the team and find time to add more data points. I am also interested in using a Horizontal Pod Autoscaler in our provisioning layer within keycloak-benchmark @ryanemerson is right about stating this Scaling guide(s) applies to beyond the scope of Cross-Site, but falls in the umbrella of Highly Available Keycloak.