Sentinela: is the application behaving correctly?

Sentinela is open-source application-logic monitoring in Python. It detects application rule violations and data inconsistencies that need correlating data across databases/APIs, applying custom logic, and tracking each affected entity until resolved.

If the problem can be expressed as "a Python function returning a list of failing entities" — via SQL, API calls, files, or any Python code joining multiple sources — Sentinela fits.

Typical cases: stuck orders (paid but never shipped), missing invoices, double charges, failed reconciliation, invalid registration data, state-transition violations.

Traditional observability (Prometheus / Grafana / Datadog) answers "is the system healthy?" (CPU, latency, error rates). Sentinela answers "is the application correct?" (per order / user / transaction). Most setups run both. See When to use Sentinela.

How it works

Write three Python functions per monitor — Sentinela schedules, tracks, alerts, and auto-resolves: 1. search() — find current issues (e.g. orders stuck in awaiting_delivery while shipment is completed). 2. update(issues) — refresh data for active issues by ID (fast, keyed lookup). 3. is_solved(issue) — return True when the entity is back to normal.

Each issue is one entity: one order, one user, one transaction. Issues roll up into alerts. Priority escalates P5 → P1 based on the rule you choose: oldest issue age, issue count, or a numeric field in the issue. You can acknowledge an alert (mark as seen) or lock it (freeze it so new issues open a fresh alert). Notifications go out via plugins (Slack, ntfy, custom).

Start with Overview, then Building a Monitor and Monitor lifecycle.

Search terms for this category

Application invariant monitoring, application logic monitoring, business invariant monitoring, business process monitoring, data quality monitoring, database consistency monitoring, entity-level issue tracking, cross-system validation, automated reconciliation, state-machine monitoring.