Development Server - HA deployment of Kibana

[kobelb]

It's incredibly common for us to recommend that users deploy multiple instances of Kibana behind a load-balancer. This allows users to get more throughput out of Kibana, and not experience downtime if a single instance crashes. As such, we must be confident that Kibana behaves properly in this configuration.

Currently, the development server implements a light-weight "base-path proxy" to ensure that users take the base path of the HTTP server into consideration. It'd be useful to expand upon this and have the development server start multiple instances of Kibana and distribute HTTP requests to them in a round-robin fashion. This will increase the likelihood that developers realize that their features aren't built with HA deployments in mind early into the development lifecycle and adjust their approach accordingly.

Ideally, this would be the default configuration, similar to how Kibana is hosted behind a base-path by default. However, if we determine that the resource utilization and developer experience is too degraded, I think it'd be reasonable to make developers opt-in to this behavior. This will still significantly decrease the burden of manually configuring their development environment to run Kibana in a HA configuration.

[elasticmachine]

Pinging @elastic/kibana-operations (Team:Operations)

[tylersmalley]

I think this would get pretty complex when running in development, but also agree we should have testing coverage for it. What do we think about changing the focus for this to support it in the FTR?

[tylersmalley]

Unfortunately, this is not something we can prioritize in the foreseeable future.