Bergamot Deployment Guide
Bergamot Monitoring is split into a number of services which needs to be deployed.
The core cluster services: Bergamot UI, Bergamot Notifier, PostgreSQL and ZooKeeper make up the central cluster which managed scheduling and processing.
Multiple workers can be deployed both local to the cluster and remote.
All UI nodes which form the coherrent UI cluster should be deployed into the same geographic site. The clustered UI nodes use HazelCast to share cached objects and state between them. As such the UI nodes will communicate heavily between each other and should have fast, low latency networking connections between them.
The UI nodes will also require a fast, low latency connection to the PostgreSQL database servers. For most use caes a 1Gb/s ethernet network will be suitable.
When deploying more than one UI node, you will need to deploy a load balancer in front of the UI nodes. This load balancer must support sticky sessions, a user should always be toggled to a single backend for as long as possible. Note that the UI nodes also use HazelCast to share session state, however requests should not flip flop between servers.
The UI nodes require to be able to connect to the ZooKeeper cluster and the PostgreSQL database servers.
You should deploy multiple worker nodes into each geographic site that you have. This is to offer load balancing of check executions as well as redundancy.
Each geographic site should be allocated it's own worker pool name, this will be used to route check executions to a particular site. The workers deployed into a geographic should all be configured with the same worker pool name.