Area of expertise
OpenTelemetry and observability
Collecting more telemetry does not automatically shorten investigations. Without operational questions, conventions, and volume control, signals become expensive yet remain hard to interpret.
Operational definition
Observability connects traces, metrics, and logs to reliability questions. OpenTelemetry provides instrumentation conventions and pipelines, but does not replace the service model or operational decisions.
Problems actually encountered
-
01
Abundant telemetry without a service model
-
02
Cardinality and retention causing unexpected cost
-
03
Context propagation broken across components
-
04
Dashboards unable to explain an incident
Decisions involved
- Define the questions each signal must answer
- Choose conventions, instrumentation points, and sampling strategy
- Balance pipeline, storage, retention, and provider independence
Common errors
- Instrumenting everything before defining use cases
- Copying dashboards without checking semantics
- Moving provider dependency into an opaque collection layer
Signals that external expertise becomes useful
- Every incident triggers ad hoc collection
- Costs rise faster than useful coverage
- Two teams assign different meaning to the same attribute
Omnivya method
-
1.
Start from representative incidents and decisions
-
2.
Map services, propagation, and signal consumers
-
3.
Assess telemetry quality, cost, and gaps
-
4.
Define conventions and testable changes
Possible deliverables
- Service model and signal-question matrix
- Documented OpenTelemetry conventions
- Instrumentation, sampling, and cost-control plan
Evidence of reasoning
- Repeatable investigation queries
- Conventions validated on real flows
- Telemetry budget tied to use cases
Explicit limits
- No recommendation to retain everything
- No presenting OpenTelemetry as a complete strategy
- No guarantee that a tool will resolve incidents
Related mission formats
- Targeted diagnosis
Clarify one precise technical decision from a bounded set of observations.
- Platform audit
Assess what a platform actually enables teams to decide and operate.
- Architecture review
Have an architecture decision reviewed before it becomes irreversible.
Frequently asked questions
- Does OpenTelemetry reduce provider lock-in?
- It improves instrumentation portability, but storage, queries, dashboards, and operating practices may remain dependent.
- Should we start with traces?
- Not always. The first signal depends on the question, system, and quality of existing data.
- Can you review observability costs?
- Yes, by connecting ingestion, cardinality, retention, and queries to the use cases that justify those costs.
Make observability testable
If this tension is yours, let’s frame the decision before widening the scope.