The usual open-source answer to Datadog is the Grafana stack: Prometheus for metrics, Loki for logs, and Grafana for dashboards and alerts, with Tempo added for traces. Together they cover Datadog's core observability jobs on infrastructure you control, and never bill per host or per data series. In return, you run the stack yourself.
Short answer
Run Grafana with Prometheus and Loki. Prometheus scrapes metrics, Loki stores logs, and Grafana puts both on dashboards with alerting on top. All three are open source and self-hostable, so your telemetry stays on your servers. Add Tempo for traces and OpenTelemetry for instrumentation to cover metrics, logs and traces as Datadog does.
This is the observability stack our team deploys and runs. I am Manoj, at MithTech, a team of Frappe specialists in Bengaluru; our small, senior team of 35 uses AI throughout delivery, and monitoring the systems we host is part of the job.
The Grafana stack matches Datadog on capability, not on convenience. You trade one polished vendor for parts you assemble and maintain yourself.
What replaces Datadog in open source?
Grafana with Prometheus and Loki. Prometheus collects and stores metrics, Loki collects and stores logs, and Grafana shows both and fires alerts. Each project is actively maintained and runs at large scale in production.
Datadog bundles infrastructure monitoring, log management, APM and more into one managed platform, and that convenience is what you pay for. The Grafana stack splits the same jobs into focused tools that work together. You lose the single-pane-out-of-the-box experience and gain your data plus a licence with no meter.
8 rows · click a column to sort
| Licence | Proprietary SaaS, usage-billed | Open source, free to self-host |
| Metrics | Built in | Prometheus |
| Logs | Built in | Loki |
| Traces / APM | Built in | Tempo + OpenTelemetry |
| Dashboards | Built in | Grafana |
| Alerting | Built in | Grafana alerting / Prometheus Alertmanager |
| Data location | Datadog's cloud | Your infrastructure |
| Who runs it | Datadog | You, or whoever runs it for you |
How do the pieces map to Datadog's features?
Each Datadog pillar has its own project. Prometheus scrapes numeric metrics from your services on a schedule and stores them in a time-series database you query with PromQL. Loki ingests log streams and indexes them by labels rather than full text, which keeps storage cheap. Grafana sits on top and turns both into dashboards and alerts.
The glue is consistent labels. When Prometheus metrics and Loki logs share a label scheme, you can jump from a spiking metric to the logs behind it in one Grafana view, which is the correlation people value in Datadog.
- Metrics: Prometheus, with exporters for databases, queues, hosts and Kubernetes.
- Logs: Loki, built to be "like Prometheus, but for logs": label-indexed and cheap to store.
- Dashboards: Grafana, with a large panel library and many data sources beyond these two.
- Alerts: Grafana alerting or Prometheus Alertmanager, routing to email, Slack, PagerDuty and others.
- Traces: Tempo, fed by OpenTelemetry instrumentation in your applications.
Is the Grafana stack really free?
The core projects are open source under recognised licences, and self-hosting them has no licence fee. Grafana, Loki and Tempo are AGPL-3.0; Grafana Labs relicensed them from Apache 2.0 in 2021. Prometheus is Apache 2.0 and was the second project to graduate from the CNCF, after Kubernetes.
Free to self-host is not zero cost. You pay for servers, storage and the engineering time to run them. What you avoid is a per-host or per-volume subscription that grows with your footprint. For teams whose telemetry grows faster than revenue, that is the point.
What does Datadog do that the open-source stack does not?
Datadog's advantage is that everything is integrated, hosted and supported on day one: APM, real-user monitoring, synthetic tests, security monitoring and hundreds of vendor-maintained integrations under one login, with a support contract behind it. Building the same breadth from open-source parts takes deliberate work.
The Grafana stack closes most of that gap if you do the assembly. Traces need Tempo plus OpenTelemetry. Long-term metric storage at large scale may need Mimir or Thanos. Upgrades, high availability, retention, backups and access control are yours to plan.
| Time to first dashboard | Datadog is faster to stand up |
| Breadth on day one | Datadog bundles more in one place |
| Traces / APM | Grafana needs Tempo + OpenTelemetry added |
| Metric storage at scale | May need Mimir or Thanos |
| Ongoing operations | Yours in the self-hosted stack |
| Data ownership and residency | Self-hosting wins |
Who should self-host observability?
Teams for whom data ownership, residency or long-term cost matter more than day-one convenience, and who have, or can hire, someone to run the stack. Teams on their own servers, teams with data-location requirements, and teams watching a usage-based monitoring bill climb are the natural fit. So are teams already on Kubernetes, where Prometheus is the standard metrics system.
Stay on managed SaaS if your team is small, observability is far from your core work, and the subscription is affordable at your scale. The right answer is the stack your team can actually run through an incident at 2 a.m.
Instrument with OpenTelemetry
Emit metrics, logs and traces from your applications with OpenTelemetry, the vendor-neutral standard, so you are not tied to any one backend.
Collect metrics with Prometheus
Deploy Prometheus to scrape your services and hosts, and add exporters for your databases and queues.
Collect logs with Loki
Ship logs into Loki with the same label scheme as your metrics, so the two correlate.
Visualise and alert in Grafana
Point Grafana at Prometheus and Loki, build dashboards, and route alerts to your on-call channels.
Add traces and plan for scale
Add Tempo for distributed tracing, and look at Mimir or Thanos for long-term, high-volume metric storage as you grow.
Frequently asked questions
+Is Grafana a full replacement for Datadog?
Grafana alone is the dashboard and alerting layer. With Prometheus for metrics, Loki for logs, and Tempo plus OpenTelemetry for traces, the combined stack covers Datadog's core jobs. You assemble and run the pieces yourself instead of buying one managed platform.
+Is the Grafana observability stack free?
The core projects are free to self-host: Grafana, Loki and Tempo are AGPL-3.0, Prometheus is Apache 2.0. You still pay for servers, storage and the time to run them. It is a different cost model, not zero cost.
+Prometheus or Datadog for a Kubernetes cluster?
Prometheus is the standard open-source metrics system for Kubernetes, with native service discovery and a large set of exporters. If you can run the stack, Prometheus with Grafana is the natural fit. Datadog stays attractive if you would rather pay for a fully managed experience.
+Can I keep all my monitoring data in India?
Yes; that is a main reason to self-host. Prometheus, Loki and Grafana run on infrastructure you choose, so every metric and log line stays where you put those servers. With a SaaS platform, telemetry sits in the vendor's cloud regions.
+What about traces and APM?
Traces come from Grafana Tempo, fed by OpenTelemetry instrumentation in your applications. That gives you distributed tracing comparable to Datadog APM. It is an extra component to deploy and instrument, so most teams add it once metrics and logs are stable.
+Do I need a team to run this myself?
You need someone who owns it, not necessarily a large team. One capable engineer can run a small Grafana stack, but upgrades and incident response take real time. If you would rather not staff that, someone can host and maintain it for you while you keep the data.
If you would rather self-host your monitoring than rent it, we can stand up and run a Grafana, Prometheus and Loki stack for you as part of managed hosting, alongside the Frappe and ERPNext systems we run. The directory of open-source SaaS alternatives covers the rest of the stack.
Comments & ratings
No comments yet. Start a new discussion.
Manoj · Published on 20 July 2026
Worked with us? Leave a Google review