This section documents the B Brands monitoring platform — the single layer
that watches every app and integration across
development, certification
and production, publishes a public status page per environment, and is
defined entirely as code (Terraform) and operated through GitHub Actions.What this is
Every environment has its own Better Stack workspace with a public status page. Instead of a human deciding whether a component is healthy, the status is derived from real production traffic flowing throughhorizon-api: the API
computes a health verdict per component, Better Stack polls it, and the status
page reflects reality automatically.
Two independent Better Stack products are used:
Uptime
Status pages, HTTP monitors and cron heartbeats. Drives the green / yellow /
red state each visitor sees.
Telemetry
Log sources, metrics and log-based alerts. Powers infrastructure charts and
the (optional) per-service error-rate alerts.
The three environments
Each environment is fully isolated: its own status page, its own monitors, its own API host and its own tokens. Nothing is shared across environments.What each color means
The status page speaks to two audiences at once — the business (is the platform usable?) and engineering (what exactly is failing?).Operational (green)
The component serves traffic normally. Error rate is within the healthy
band.
Degraded (yellow)
The component still responds but a sustained error rate, a failing route or
a stale cron was detected. Users may notice slowness or partial failures.
Down (red)
A real outage: the short-window error rate crossed the outage threshold. The
component endpoint returns
503.What is monitored
Every environment converges to the same catalogue, each pointing at its own hosts.30 HTTP monitors
10 frontend apps + 11 API categories + 9 integrations. Better Stack polls
each one every 1-3 minutes.
3 cron heartbeats
Health API, Health Integrations and Metrics Push. Confirm the scheduler
itself keeps running.
4 status-page sections
Platforms, API, Integrations and Infrastructure — how components are grouped
for visitors.
Telemetry sources
One log/metrics source per environment for infrastructure charts and log
alerts.
How the signal flows
The golden rule: monitors decide red, Status Reports decide yellow
This is the single most important concept for the whole platform. The two health layers are deliberately independent so that a scheduler problem can never fake an outage.Red (down) — HTTP monitors
A monitor polls the component endpoint. It only sees
503 when the
short-window error rate is a real outage. This path has no dependency on
the cron scheduler, so a missed tick cannot produce a false red.Yellow (degraded) — Status Reports
The health crons compare each verdict against the published state and
create/resolve Status Reports. A missed tick only delays the yellow
transition, which is harmless.
Where to go next
Health Architecture
How the tri-state verdict is computed and published.
Infrastructure as Code
The Terraform definition of every Better Stack resource, and
plan vs
apply.CI/CD for Monitoring
The GitHub Actions workflow that validates and applies changes safely.
Operations Runbook
Day-2 tasks: sync env vars, add a monitor, rotate tokens, resolve drift.