We’re not on Kubernetes. Does this still work?

Yes. Run Prometheus and the OTel Collector on VMs, scrape via static configs, and use your existing deploy tooling. Argo Rollouts becomes a canary proxy or feature flag system (e.g., Flagger, LaunchDarkly) with the same SLO checks.

What if we can’t change the app code quickly?

Start with sidecar/proxy metrics (Envoy/Nginx ingress), synthetics, and blackbox probes. Add tracing via auto-instrumentation where possible. You’ll get 60–70% of the value before touching core code.

How many SLOs should we have?

One to two per critical user journey. More than five per service becomes dashboard cosplay. Expand only when you have budget reviews running smoothly.

How do we pick good thresholds?

Use historical percentiles and business feedback. If 95% of checkouts complete under 700ms and users complain above 1s, pick 800ms for a 99% objective. Revisit quarterly.

Guides · Oct 15, 2025 · 10 minute read

The 30‑Day Hardening Plan for a Legacy Service: Progressive Observability and SLOs That Stick

No rewrites. No Big Bang platform migrations. Just a month of disciplined, incremental moves that stop the bleeding and keep you shipping.

Riley Grant

Principal Engineer, GitPlumbers

20 years hardening systems at scale. Ex-SRE lead at a fintech, platform re-architecture at a top-10 retailer, and early customer-zero of OpenTelemetry. I fix pager fatigue and ship safely.

If it isn’t measured, it’s folklore. If it doesn’t gate a release, it’s a wish.

Back to all posts

The 30‑Day Hardening Plan for a Legacy Service: Progressive Observability and SLOs That Stick

Key takeaways

Implementation checklist