Internal demonstrator

SiteGuard Monitor: service-health and alerting demonstrator

We implemented Python/SQLite probes, durable health observations and a status view, then exercised the alert states with synthetic failure sequences.

DIAGNOSIS PATH / PRODUCTIONFollow the failing request

Verify rollback and recovery

Trace, correct and rerun the failing operation.

RESOURCES / ENGINEERING STORIES

The problem, the change and the evidence.

An internal Python and SQLite monitoring prototype using simulated service-health checks.

The implementation records health observations and exposes the status transitions needed to investigate a failed check.

Explore the system, engineering decisions and observed outcome in the sections below.

Project context: An internal Python and SQLite monitoring prototype using simulated service-health checks.

Architecture and trust boundaries

Internal demonstrator — synthetic service-health observations only.

Failures and timestamps are synthetic; alert delivery, real multi-region probes and SLO accuracy were not validated.

Repeat deterministic success/failure inputs and inspect stored status transitions. No production customer is implied.

A successful HTTP request alone does not establish a healthy dependency or an actionable incident signal.
Operators need enough history to distinguish a brief outage from recurring failures or a noisy check.

Developed a local probe workflow for checking service responses and recording observations in SQLite.
Exposed uptime history and alert-oriented status in a small demo interface.
Kept synthetic failure conditions separate from genuine production traffic so tests were repeatable.

Used synthetic success and failure responses to exercise state transitions and stored observations.
Checked that observed failures could be surfaced for inspection and correlated with the recorded timeline.
For any production adoption, additional tests would be required for retries, clock skew, false positives, notification delivery and alert fatigue.

The implementation records health observations and exposes the status transitions needed to investigate a failed check.
The local prototype was tested with simulated failures; alert delivery and production reliability would require additional validation.