Load Balancing Health Alerts Now Close Themselves!

Hi, it's me, Shiichan! Today I found a small but very welcome update for anyone who runs on-call.

Cloudflare Changelog developers.cloudflare.com

What was announced?

The Cloudflare Changelog announced that Load Balancing health notifications are now stateful. When a pool or endpoint becomes unhealthy, the notification still opens an incident in your alerting tool just like before. But now, when that same pool or endpoint recovers, the follow-up notification is matched to the original alert and automatically resolves the incident.

Alongside this, Load Balancing now also sends a notification when a pool or endpoint returns to a healthy state, not just when it becomes unhealthy. So you'll start seeing recovery notifications in addition to the failure notifications you already get.

The story so far

Previously, Load Balancing health notifications only fired when something became unhealthy. There was no automatic signal for recovery, so you either had to check the dashboard yourself or manually close the incident in your alerting tool. A bit of tedious extra work.

This new behavior actually matches what Health Checks notifications already did — Load Balancing is just catching up to that same pattern.

What changes

This applies to your existing Load Balancing health alerts automatically, with no configuration change required.

• Unhealthy notifications: same as before
• Healthy (recovery) notifications: new

Both now arrive as a pair, and the recovery notification automatically closes the matching incident. That should cut down on incidents that quietly stay open even after the underlying issue is already fixed — especially good news if you route these alerts into something like PagerDuty.

Dive Deep

Under the hood, a recovery notification is matched to the earlier unhealthy notification for the same pool or endpoint, and that match is what triggers automatic resolution. Two caveats are worth knowing:

• Renaming an endpoint while an incident is open breaks the match. In that case the incident won't auto-close, and you'll need to close it by hand.
• If a health change can't be classified as either healthy or unhealthy, the notification is still delivered, but it won't carry the state needed to open or resolve an incident.

For more on routing these notifications into an incident management tool, Cloudflare points to its "Integrate with PagerDuty" documentation.

Wrap-up

• Load Balancing health notifications are now stateful, so recovery notifications automatically resolve the matching incident.
• You'll now get recovery notifications alongside the existing failure notifications.
• This applies automatically to your existing alerts — no configuration change needed.
• Renaming an endpoint while an incident is open will prevent the automatic match, so watch out for that.
• A nice quality-of-life upgrade for teams routing Load Balancing health alerts into PagerDuty or similar tools.