Logs, metrics and traces on Railway without Datadog
Short answer: deploy a self-hosted observability stack next to your app and send it OpenTelemetry data. Two Tracklayer templates do this on Railway. The Grafana LGTM template runs Grafana, Loki (logs), Tempo (traces), Prometheus (metrics) and Alloy as one OpenTelemetry endpoint. The SigNoz template runs SigNoz with ClickHouse as a single application for traces, metrics and logs. Both keep data in your own Railway project.
Deploy Grafana LGTM Deploy SigNozWhen this makes sense
Platform logs show what one service printed. They do not show a slow request across an API, a worker and a queue. Traces, metrics and logs together answer that. Hosted APM tools charge per host or per seat. Self-hosting trades that bill for resource usage on Railway and for operating the stack yourself.
Observability templates people deploy on Railway
| Template | Creator | Deployments | Services |
|---|---|---|---|
| Uptime Kuma | Brody's Projects | 1,456 | 1 |
| Langfuse v3 | Langfuse's Projects | 688 | 6 |
| Grafana 13 LGTM Observability Stack | Tracklayer | 10 | 5 |
| SigNoz APM for Railway Apps | Tracklayer | 3 | 5 |
| Helicone LLM Observability Gateway | Tracklayer | 2 | 7 |
| Langfuse v4 LLM Observability Stack | Tracklayer | 2 | 5 |
Railway marketplace data as of 2026-11-19. Full ranking: Best observability and monitoring templates on Railway.
The list mixes general monitoring tools with LLM-specific ones. For LLM call tracing, see the Langfuse guide.
Which one to choose
| Grafana LGTM | SigNoz | |
|---|---|---|
| Shape | Separate tools for dashboards, logs, traces and metrics, linked in Grafana | One application with a ClickHouse store |
| Services | Grafana, Alloy, Loki, Tempo, Prometheus | SigNoz, OTel Collector, schema migrator, ClickHouse, ClickHouse Keeper |
| Storage | Volumes plus two Railway Buckets (logs, traces) | Volumes, mainly ClickHouse |
| Strengths | Linked datasources, service graph, Prometheus scraping of /metrics |
Request rate, errors and duration per service, alerts by email, Slack or webhook |
If you already know Grafana dashboards or want to scrape Prometheus endpoints, choose LGTM. If you want one UI for application performance monitoring, choose SigNoz. Running both is possible but doubles the cost.
Deploy and connect, step by step
- Use Hobby or higher. Both templates give Hobby as the minimum.
- Deploy a template. Grafana LGTM needs no input. SigNoz asks for your email in
SIGNOZ_USER_ROOT_EMAIL. - Wait for healthy services. SigNoz takes a few minutes on first boot while the schema is created.
- Log in as described in the FAQ.
- Connect an app in the same project. Set these variables on the service you want to observe.
For Grafana LGTM:
OTEL_EXPORTER_OTLP_ENDPOINT=${{alloy.OTEL_EXPORTER_OTLP_ENDPOINT}}
OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
OTEL_SERVICE_NAME=your-service-name
For SigNoz:
OTEL_EXPORTER_OTLP_ENDPOINT=http://${{otel-collector.RAILWAY_PRIVATE_DOMAIN}}:4318
OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
OTEL_SERVICE_NAME=${{RAILWAY_SERVICE_NAME}}
Then start the app with your language’s OpenTelemetry auto-instrumentation. Port 4317 serves gRPC on both. From outside Railway, use the public domain and the bearer token (OTLP_PUBLIC_TOKEN).
In Grafana, open Explore and pick Prometheus, Loki or Tempo; the stack’s own metrics appear from the start. Grafana LGTM can also scrape /metrics endpoints through EXTRA_SCRAPE_TARGETS on the prometheus service.
What it costs
Railway bills usage. Idle RAM for Grafana LGTM was measured locally by the template author; the other figures are estimates, calculated at $10 per GB of RAM, $20 per vCPU, $0.15 per GB of volume and, for Grafana, $0.015 per GB of bucket per month.
| Stack | Load | RAM | Estimated monthly cost |
|---|---|---|---|
| Grafana LGTM | Idle (RAM measured) | about 0.6 GB | about $7 |
| Grafana LGTM | 5 to 15 instrumented services (estimated) | about 2.5 GB, about 30 GB in buckets | about $35 |
| SigNoz | Idle (estimated) | about 0.95 GB | about $12 |
| SigNoz | Small team, few services, 7-day retention (estimated) | about 2.5 to 3.8 GB | about $40 to $50 |
Ingest volume drives memory for Loki, Tempo and ClickHouse. Retention drives bucket and volume size. SigNoz heavy-ingest estimates reach about $100 per month.
Limitations and licence notes
- Single node. Grafana LGTM is a single-node stack that scales vertically, and Grafana uses SQLite, so it stays at one replica. SigNoz’s collector is stateless and can add replicas; ClickHouse needs more memory for heavier queries.
- Storage limits. Hobby volumes are limited to 5 GB per service. Set retention to fit, or move to Pro.
- Upgrades. Versions are pinned. For SigNoz, change the
signozandschema-migratortags and the collector’sFROMtag together and redeploy the migrator first. For Grafana LGTM, change theFROMtag per service and read release notes first, since Loki and Tempo major versions can change configuration. - Licences. See the FAQ.
- Community templates. Not affiliated with or endorsed by Grafana Labs, the Prometheus project, SigNoz Inc. or ClickHouse Inc.
Browse more guides.
FAQ
How do I log in to Grafana or SigNoz?
For Grafana, open the grafana domain and sign in as admin with GF_SECURITY_ADMIN_PASSWORD from the grafana variables; it is applied on first start. For SigNoz, enter your email in SIGNOZ_USER_ROOT_EMAIL when deploying and sign in with the generated SIGNOZ_USER_ROOT_PASSWORD from the signoz variables once the stack is healthy.
How long is data kept?
The Grafana stack defaults to 31 days for Loki logs, 14 days for Tempo traces and 15 days for Prometheus metrics, all adjustable by variables. SigNoz retention is set in Settings, General; the template notes suggest 3 to 7 days on Hobby because of its 5 GB volume limit.
Can apps outside Railway send data?
Yes. Both templates expose a token-protected public OTLP endpoint. Requests without the bearer token are rejected.
What are the licence terms?
According to the NOTICE files, Grafana, Loki and Tempo are AGPL-3.0, and Alloy and Prometheus are Apache-2.0. For SigNoz, the server image contains MIT code plus code under the SigNoz Enterprise License that stays disabled without a licence key; its collector is AGPL-3.0 and ClickHouse is Apache-2.0. The AGPL matters mainly if you modify the software and offer it to others over a network. Seek legal advice if unsure.
Do I need a paid Railway plan?
Both templates list Hobby as the minimum plan. Pro is suggested once telemetry exceeds the 5 GB volume limit per service.