Tame a flapping monitor
You’ll get: an explanation of the notification storm InfraInbox already prevents on its own when a check flips repeatedly, and what to adjust when the underlying check itself needs fixing.
What InfraInbox already does
Section titled “What InfraInbox already does”Nothing to set up here — this is automatic. If an incident reopens three times within 30 minutes, InfraInbox marks it as flapping: it sends one notice saying so, and holds back the open/resolved notices each further flip would otherwise cause, so you don’t get paged a dozen times for one unstable check. Once the incident has gone 30 minutes without another reopen or resolve, InfraInbox sends one final notice and treats it as settled again.
The incident itself keeps updating in the inbox and its activity timeline the whole time — flapping only quiets the notifications, not the record of what happened.
When to do something about it
Section titled “When to do something about it”Flapping is a symptom, not a setting: it means the thing you’re watching is genuinely unstable, or your check is more sensitive than the thing it’s watching deserves. A few common causes and fixes:
- A heartbeat’s grace is too tight. If a host or job monitor flips between up and down because it’s occasionally a little slow rather than actually failing, widen its Grace before alerting on the monitor’s edit dialog — see Heartbeat monitors.
- The upstream check itself is flapping (a Proxmox VM briefly reporting unreachable during a live migration, an Uptime Kuma check with too short a retry interval). Fix the check’s own sensitivity at the source, if it has one — InfraInbox is reporting what it’s told.
- You know it’s going to be unstable for a while (a known-flaky service you’re actively working on). Use a time-boxed silence instead of waiting for automatic flapping detection to kick in — see Silence alerts during maintenance.
Check it works
Section titled “Check it works”Once the underlying cause is fixed, the incident stops reopening and resolves normally — no more “flapping” notices for it. If you widened a heartbeat’s grace, a normal, brief slowdown no longer opens an incident at all; watch the monitor’s state on the source’s Heartbeats tab for a few cycles to confirm.
- How incidents work — the full incident lifecycle, flapping included.
- Heartbeat monitors — interval and grace, in full.