How we reduce alert noise with dependency-aware checks

Alert fatigue is one of the most common complaints from on-call engineers. When an upstream service goes down, every downstream monitor fires at once — and the team wakes up to dozens of notifications about symptoms rather than one notification about the root cause.

StackEye addresses this with dependency-aware checks. You can declare that a set of monitors depend on a shared upstream (a load balancer, a CDN origin, or a database cluster). When the upstream fails, StackEye suppresses the downstream alerts and sends a single grouped incident instead.

How grouping works

Monitors in a group share a common parent check. The regional probe runs all checks independently, but the API server applies fan-out suppression before dispatching notifications. If more than a configurable threshold of monitors in a group fail within the same check window, a group-level incident is created and individual alerts are silenced.

This mirrors how experienced incident responders think: identify the blast radius first, then drill into components.

Configuring groups

Groups are created in the StackEye dashboard under Probe Groups. You assign monitors to a group and optionally specify a parent probe. The parent probe is checked first; if it fails, the group is immediately marked affected and downstream checks are skipped for that cycle.

group: production-api
  parent: api-gateway-health
  members:
    - /v1/users
    - /v1/payments
    - /v1/auth

Results in practice

Teams using group-based monitoring report a 60–80% reduction in per-incident notification volume during cascading failures. The remaining alerts are signal, not noise.

Dependency-aware monitoring is available on all paid plans. Free plan users can create one group with up to five members.