Deploy and Host OpenObserve on Railway

OpenObserve is an open-source observability platform for logs, metrics, traces, and real user monitoring (RUM) — a Rust-native, Parquet-backed alternative to Elasticsearch, Datadog, and Splunk that the makers claim ingests at up to 140× lower storage cost than ELK. Self-host OpenObserve when you want the full observability stack (search, dashboards, alerts, pipelines, RUM) behind a single binary and a single web UI.

This Railway template runs OpenObserve in single-node local mode from the official openobserve/openobserve Docker image, mounts a persistent volume at /data for the embedded SQLite meta-store and Parquet shards, and exposes the web UI on HTTPS at a generated .up.railway.app domain. No external Postgres, Elasticsearch, or S3 bucket is required — local mode is feature-complete.

Getting Started with OpenObserve on Railway

After the deploy goes green, open the generated https://*.up.railway.app URL — you will be redirected to /web/ and prompted to log in. Use the email and password you set in ZO_ROOT_USER_EMAIL and ZO_ROOT_USER_PASSWORD. Once inside, the default organization is default; create your first stream (Streams → Add Stream) to start ingesting logs, then point any OpenTelemetry collector, Fluent Bit, Vector, or curl/JSON producer at /api/default//_json with HTTP basic auth. Dashboards, alerts, and pipelines are all available from the left nav. Change the root password from User → Settings before sharing the URL — it was set once at first boot and is not rotated by redeploys.

About Hosting OpenObserve

OpenObserve is a unified observability stack: a single Rust binary that ingests, indexes, and queries logs, metrics, traces, and frontend RUM events using Apache Parquet columnar storage and DataFusion for SQL/PromQL queries. The community edition is AGPL-3.0 and feature-complete — same UI, same query engine, same pipelines as the cloud offering.

Key features:

• Logs, metrics, traces, and RUM in one binary — no Elasticsearch, no separate metrics TSDB
• Parquet on disk + DataFusion — ~140× cheaper storage than ELK, claimed by the project
• Built-in dashboards, alerts, pipelines (VRL/JS transforms), and a query language for each signal
• OTLP, Prometheus, Loki, Elasticsearch _bulk, and Fluent Bit ingest APIs out of the box
• Built-in user/auth/RBAC and multi-tenant organizations

Why Deploy OpenObserve on Railway

Railway is the fastest way to put a single-binary OpenObserve in front of real traffic:

• One-click deploy of openobserve/openobserve — no Dockerfile, no Helm chart
• Persistent volume at /data so SQLite + Parquet survive redeploys
• HTTPS, custom domain, and zero-config private networking included
• Logs, metrics, and deploy controls in the same dashboard you already use

Common Use Cases

• Centralised log aggregation for self-hosted services running on Railway, Kubernetes, or bare-metal
• Metrics + traces backend for OpenTelemetry-instrumented apps (Prometheus and OTLP compatible)
• Drop-in Elasticsearch _bulk endpoint for existing Filebeat/Logstash/Fluent Bit pipelines
• Frontend RUM (real user monitoring) for SPAs without paying per-session Datadog/Sentry fees

Dependencies for OpenObserve

• OpenObserve — openobserve/openobserve:v0.80.3 (single Rust binary, AGPL-3.0)

Environment Variables Reference

Deployment Dependencies

• Runtime: Docker image openobserve/openobserve (Docker Hub)
• Source: github.com/openobserve/openobserve
• Docs: openobserve.ai/docs
• Environment variable reference: openobserve.ai/docs/environment-variables

Hardware Requirements for Self-Hosting OpenObserve

These numbers cover the single-node local-mode template. Real disk usage scales with ingest rate — Parquet compression typically lands between 8× and 40× of the raw payload.

Self-Hosting OpenObserve

Minimal Docker invocation that mirrors the Railway template:

docker run -d \
  -p 5080:5080 \
  -v openobserve_data:/data \
  -e [email protected] \
  -e ZO_ROOT_USER_PASSWORD=ChangeMe123! \
  -e ZO_DATA_DIR=/data \
  -e ZO_LOCAL_MODE=true \
  -e ZO_TELEMETRY=false \
  openobserve/openobserve:v0.80.3

Ingest a log line via the Elasticsearch-compatible _json endpoint:

curl -u [email protected]:ChangeMe123! \
  -H 'Content-Type: application/json' \
  -d '[{"level":"info","msg":"hello openobserve","service":"demo"}]' \
  http://localhost:5080/api/default/demo_logs/_json

Once it returns {"code":200,...}, navigate to Streams → demo_logs in the UI to see the row.

How Much Does OpenObserve Cost to Self-Host?

OpenObserve's community edition is open-source under AGPL-3.0 — the binary, the UI, dashboards, alerts, pipelines, and RUM are all free. The only spend is Railway infrastructure (CPU, RAM, volume storage, egress); a 512 MB / 1 vCPU instance with a small volume runs well within Railway's Hobby plan. The hosted OpenObserve Cloud and Enterprise tiers are billed by ingest volume — by self-hosting you swap that for Railway's flat per-resource pricing, which is typically 5–20× cheaper at moderate ingest rates.

OpenObserve vs Elasticsearch and Datadog

OpenObserve trades Lucene's mature full-text features for Parquet's columnar economics — typical disk usage is a fraction of an equivalent ELK stack.

FAQ

What is OpenObserve and why self-host it? OpenObserve is an open-source, Rust-based observability platform (logs, metrics, traces, RUM) that uses Parquet on local disk or S3 instead of Lucene shards. Self-hosting keeps telemetry data inside your own infrastructure and avoids per-GB Datadog or Splunk bills.

What does this Railway template deploy? A single OpenObserve container (openobserve/openobserve:v0.80.3) in local mode with ZO_LOCAL_MODE=true, a persistent volume mounted at /data, and the web UI exposed on a generated HTTPS domain.

Why doesn't this template include Postgres or S3? OpenObserve local mode uses an embedded SQLite meta-store and local-disk Parquet — no external dependencies. Postgres and S3 are only required for cluster mode (multiple ingester/querier/router/compactor nodes), which is out of scope for a single-service template.

Can I change the admin password after deploy on Railway? Change it from inside the OpenObserve UI under User → Settings. The ZO_ROOT_USER_PASSWORD env var is read only at first boot — editing it on Railway after the first deploy will not rotate the password and may lock you out if the volume is reset.

How do I send OpenTelemetry data to self-hosted OpenObserve? Point any OTLP-compatible collector at https:///api/default/ with HTTP basic auth using the root credentials, or generate a per-stream ingest token from Settings → Ingestion. OpenObserve accepts OTLP/HTTP for logs, metrics, and traces.

Do I need the latest-simd image instead of v0.80.3? Only if your Railway region runs on x86_64 CPUs with AVX512 (or ARM with NEON), and you ingest enough data to feel the SIMD speedup. For small to medium deployments the plain tag is the safer default.