Observability
Observability is the property of a system that allows its internal state to be understood from the signals it emits, including for failures nobody anticipated.
What is observability?
Observability is the property of a system that allows its internal state to be understood from the signals it emits, including for failures nobody anticipated.
Plain and technical
Monitoring tells you that something is wrong, because someone decided in advance to watch for it. Observability is being able to work out why something is wrong even when nobody predicted that particular failure.
Observability is conventionally achieved through metrics, logs and traces, though having all three is not sufficient: they must share an identity model so that a metric leads to a trace and a trace leads to the logs emitted during it. The practical test is whether a new question about production behaviour can be answered without first deploying additional instrumentation.
What it looks like in practice
Nearby vocabulary
Monitoring
Monitoring is the practice of collecting predefined signals from a system and alerting when they leave their expected ranges.
Distributed tracing
Distributed tracing follows one request across every service it touches, recording the duration and outcome of each step so you see where time went.
OpenTelemetry
OpenTelemetry is a vendor-neutral standard and set of libraries for generating and exporting metrics, logs and traces, portable across backends.
Cardinality
Cardinality is the number of distinct combinations of label values for a metric, and it is the primary driver of metric storage cost.
How DevOpsArk handles observability
Articles on this subject
What is observability, and how is it different from monitoring?
A definition of observability that does more work than "the three pillars": what property you are actually trying to obtain, and how to tell whether you have it.
Monitoring vs observability: a distinction worth keeping
The difference between monitoring and observability, why the distinction is more than marketing, and what each one is actually for.
More definitions
See these concepts in a running system
A 30-minute walkthrough against your own infrastructure rather than a slide about the theory.