Deploy and Host PostHog Hobby

PostHog Hobby is a reliable self-hosted deployment of PostHog — a complete product analytics platform with event tracking, feature flags, session recordings, funnels, and dashboards. This template ships a hobby-tier configuration pinned to a specific PostHog commit, with healthchecks on every service and a one-shot migration runner so upgrades stay clean.

About Hosting PostHog Hobby

This template runs the official PostHog application stack on Railway with 10 pre-configured services, each pinned to immutable image tags rather than latest. The setup mirrors PostHog's official docker-compose.hobby.yml but adapts it for Railway's per-service architecture: ZooKeeper is replaced with Redpanda (KRaft mode, single binary), ClickHouse ships pre-baked with PostHog's required cluster topology and UDFs, and migrations run as a separate one-shot service to avoid races on rebuild.

Pinned to PostHog sha-70c912d (2026-05-04). All cross-service connections use Railway's private network with auto-generated secrets. After the first deploy, redeploy the Web service once to pick up the CAPTURE_ENDPOINT reference (Railway's public domains populate after the first build).

Resources: requires approximately 8 GB RAM total. Railway Pro plan recommended (~$25-30/mo). Single-broker Kafka (Redpanda) is suitable for projects up to ~1M events/day.

Why Deploy PostHog Hobby?

PostHog Cloud charges per event, which scales painfully for high-traffic applications. This template gives you the full PostHog feature set at fixed Railway pricing — no per-event billing, no data sent to third parties, no row limits.

Unlike other PostHog Railway templates, this one:

• Pins every image to a specific commit SHA instead of :latest — no surprise breakage when upstream pushes a regression
• Uses upstream PostHog images (posthog/posthog, ghcr.io/posthog/posthog/capture, ghcr.io/posthog/posthog/plugin-server) — no fork dependency or bus factor risk
• Includes healthchecks on Web, Plugins, Capture, and ClickHouse — Railway routes traffic only to ready instances
• Runs migrations as a separate one-shot service — Web/Worker rebuild without re-running migrations, no blocking
• Ships a custom ClickHouse image with PostHog's required cluster topology, user profile flags, UDFs (funnel aggregations), and Kafka IDL schemas pre-baked
• Drops unnecessary services: no ZooKeeper (Redpanda KRaft replaces Kafka+ZK), no SeaweedFS (MinIO covers session recordings), no Temporal/livestream (advanced features moved to optional add-ons documented in README)

Common Use Cases

• Product analytics for SaaS startups — track signups, conversions, feature adoption, retention without sending data to Mixpanel/Amplitude/PostHog Cloud
• Feature flag rollouts — gradual feature releases, A/B tests, kill switches with sub-second flag evaluation in your private network
• Session recordings for support/UX teams — replay user sessions to debug issues or understand behavior, stored in your own MinIO bucket
• Funnel and retention analysis — built-in funnel UDFs (aggregate_funnel) for conversion analysis without extra setup
• GDPR/data residency compliance — keep all event data inside your own Railway project, no external processors
• Dogfood / internal tools — give your team a self-hosted PostHog instance for experiments without affecting production analytics

Dependencies for PostHog Hobby

PostHog Hobby bundles every dependency it needs as separate services in this template — you don't need to provision anything externally. The full stack:

• Web (posthog/posthog) — Django app serving the UI and REST API
• Worker (posthog/posthog) — Celery worker + scheduler for async tasks (reports, exports, cohort calculation)
• Plugins (ghcr.io/posthog/posthog/plugin-server) — Node.js event ingestion pipeline, transformations, hooks
• Capture (ghcr.io/posthog/posthog/capture) — Rust event capture service, public endpoint for SDKs
• Migrate (posthog/posthog) — one-shot migration runner (Postgres + ClickHouse), exits after success
• ClickHouse (ghcr.io/ikatsuba/posthog-clickhouse) — column store for events, with PostHog configs baked in
• Postgres 15 (postgres:15.12-alpine) — application metadata, users, dashboards
• Redis 7 (redis:7.2-alpine) — Celery broker and cache
• Kafka / Redpanda (redpandadata/redpanda) — event bus between Capture and Plugins (KRaft mode, no ZooKeeper)
• MinIO (minio/minio) — S3-compatible storage for session recordings and exports

Deployment Dependencies

• Railway account with Pro plan or comparable resource budget — minimum ~8 GB RAM total across the stack, ClickHouse alone needs 2 GB+
• Railway region close to your users (lower latency for Capture event ingestion)
• Optional: custom domain for the Web service if you want PostHog at analytics.yourcompany.com instead of *.up.railway.app

After deploy, follow these one-time steps:

1. Wait for Migrate to exit cleanly (Stopped status in Railway UI = success)
2. Verify Web, Plugins, Capture, ClickHouse show Healthy
3. Redeploy the Web service once to pick up the CAPTURE_ENDPOINT reference (circular dependency on first deploy)
4. Open the Web public URL and complete the PostHog setup wizard