Operations and reliability
Observability with OpenTelemetry
Metrics, logs and traces from services are collected through a collector, enriched and feed SLOs.
Fictional scenario. An example situation that illustrates the approach. It is not a client project and describes no measured results.
Business problem
A fictional platform has logs in different tools and no tracing. Incidents are investigated slowly.
Assumptions
- Services can be instrumented.
- An observability backend has been chosen.
- Teams agree on the key SLOs.
Constraints
- No personal data in telemetry.
- Volume and cost under control.
- Vendor independence.
Reference architecture
Components by stage — an illustration, not a running system and not a certified design.
01Sources
- Instrumented services
- Infrastructure
02Collection
- OpenTelemetry collector
- Filters and masking
- Sampling
03Storage
- Metrics
- Logs
- Traces
04Use
- Dashboards
- SLOs and error budgets
- Alerts
The diagram as a list
| Stage | Components |
|---|---|
| Sources | Instrumented services, Infrastructure |
| Collection | OpenTelemetry collector, Filters and masking, Sampling |
| Storage | Metrics, Logs, Traces |
| Use | Dashboards, SLOs and error budgets, Alerts |
Interfaces
| From | To | Via |
|---|---|---|
| Services | Collector | OTLP |
| Collector | Backend | Exporters |
| Backend | On-call | Alerts |
Data flow
- Services send telemetry over OTLP.
- The collector masks and samples.
- Data is stored by type.
- SLOs are computed from metrics.
- Alerts fire when the budget burns.
Security boundaries
- Personal data masked in the collector.
- TLS between services and the collector.
- Role-based access to dashboards.
Where a person approves
- New alerts to on-call.
- Changes to production sampling.
Delivery phases
01 · Assessment
Current signals and gaps.
02 · Foundation
Collector and backend.
03 · Instrumentation
The key services.
04 · SLOs
Targets, budgets and alerts.
Limitations
- It does not show business outcomes by itself.
- Legacy code can be hard to instrument.
Trade-offs
- More data — better visibility but more cost.
- Sampling — cheaper but misses cases.
Related services
Related assessments
Integration patterns used
Workshops and assessments
Facing something similar?
We start with an assessment of where you are and propose a realistic first phase.