From Percona PMM to Obsfly

• PMM is free but operationally expensive — you're running Prometheus + ClickHouse + Grafana yourself.
• ML-driven anomaly detection isn't built in.
• Multi-tenant access controls are minimal.
• AI advisory is missing entirely.

Fairness signal — useful in renewal conversations.

• Free and open source — no vendor lock-in.
• Battle-tested by hundreds of teams running MySQL at scale.
• Query Analytics view is genuinely good.

1. Step 1

   Audit current PMM resource cost

   Add up the EC2/EKS spend for Prometheus, ClickHouse, the PMM server, plus the engineer-hours spent maintaining it. That's your real baseline.

2. Step 2

   Run Obsfly in parallel

   Install the agent on hosts that already run pmm-agent. They don't conflict. Confirm metric parity for two weeks.

3. Step 3

   Migrate dashboards

   PMM uses Grafana. Obsfly imports common Postgres / MySQL dashboards out of the box. Custom panels: copy the PromQL, swap the data source.

4. Step 4

   Decommission PMM

   Tear down the PMM server. Keep your existing Prometheus if you use it for non-DB metrics.

• If you have heavy Grafana customization, plan for a 1-2 week dashboard rebuild.
• Verify retention requirements — Obsfly default is 30 days; long-term storage is per-tier.

PMM is free. Why pay?

PMM costs ~$30k/year in cloud + ops time at scale. Obsfly Team is $39/DB/mo with no ops overhead.

Can we keep our existing Grafana?

Yes. Obsfly exports a Prometheus-compatible scrape endpoint — point Grafana at it.