Back to docsRecipe

Recipe: Rolling update playbook

Ship Meridian payload updates to a fraction of your fleet, validate health, then progressively widen the rollout — all without taking a single endpoint offline.

Prerequisites

  • Meridian CLI v2.4+ installed on your build runner.
  • At least one active fleet with the canary tag.
  • Telemetry webhook configured in the Meridian dashboard.

Step 1 — Stage the artifact

Build and sign your payload as usual. Push it to the Meridian CDN with the --stage flag so it is available but not yet served to any endpoint.

meridian push payload.bin --stage --tag v3.2.1-rc1

Step 2 — Create a canary rule

Target endpoints carrying the canary tag. Start with 5% of the fleet and a 10-minute observation window.

meridian rollout create --tag canary --pct 5 --window 10m

Step 3 — Observe & validate

Watch crash rate, heartbeat latency, and license-check success rate in the telemetry dashboard. If any metric breaches its threshold, Meridian auto-pauses the rollout.

Step 4 — Widen or roll back

If the canary is clean, bump to 25%, then 100%. If not, run meridian rollout abort and the fleet reverts to the last known-good payload within seconds.

Pro tip

Pair rolling updates with the offline-grace recipe so endpoints that miss the update window stay licensed.