What Is Monitoring? CI/CD and Deployments Explained
Monitoring is the practice of collecting signals from a system and watching them against expectations, so you find out when something is wrong, ideally before users do.
Monitoring is the operational discipline of knowing the health of your systems at all times. It predates the word observability by decades and remains its most concrete day-to-day expression: dashboards, thresholds, and alerts. In delivery, monitoring spans both the running application and the pipeline that builds and ships it.
What monitoring covers
Monitoring instruments a system, collects measurements, and compares them against known-good ranges. When a measurement crosses a threshold, an alert fires. The classic targets are the so-called golden signals: latency, traffic, errors, and saturation. Monitoring is fundamentally about expectations you have defined in advance.
Black-box vs white-box monitoring
- Black-box: probing the system from outside as a user would, for example an uptime check hitting a URL. It tells you symptoms.
- White-box: reading internal signals the system exposes, such as queue depth or memory use. It tells you causes.
- Mature setups use both: black-box to detect user-visible problems, white-box to explain them.
Monitoring vs observability
These overlap but are not the same. Monitoring is the act of watching predefined signals and alerting on them. Observability is the underlying property that makes deep, unplanned investigation possible. You monitor the things you know to worry about; observability is what saves you when the failure is something you never thought to watch.
Monitoring the pipeline and the deploy
Treat CI/CD as a production system worth monitoring. Watch build success rate, queue time, and duration trends; alert when the failure rate spikes or a pipeline runs far longer than usual. During and after a deploy, monitor application health so a bad release triggers a fast, automated response rather than a customer report hours later.
Avoiding noise
The failure mode of monitoring is too many alerts. If every minor blip pages someone, real signals get lost and on-call burns out. Good monitoring alerts on symptoms users feel, ties each alert to a clear action, and tunes thresholds so the signal-to-noise ratio stays high.
Key takeaways
- Monitoring collects signals and compares them against expectations to detect problems.
- Black-box monitoring shows symptoms; white-box explains causes.
- It complements observability rather than replacing it.
- Pipelines and deploys are worth monitoring, not just the running app.