This document gives an overview of how a deployed PostHog Helm chart works on Kubernetes.
The general architecture looks as follows:
No communication is needed into or out of this namespace other than the ingress controller for app and collecting data.
Note that the specifics of this may vary:
- ClickHouse, Kafka, PostgreSQL and Redis services may be hosted outside of the namespace or configured differently.
- The ClickHouse cluster is managed by clickhouse-operator and the exact number of pods vary according to sharding settings.
Event ingestion is explained in more depth in our engineering handbook here.