What is OpenTelemetry and how does it help find errors between services?

Published on 2026-10-20 by frfrand

OpenTelemetry explicado con ejemplos: traces, spans, métricas, contexto distribuido y límites de instrumentación.

OpenTelemetry is an open project that offers APIs, SDK, tools, and conventions to produce and transport telemetry data, such as traces, metrics, and logs. It helps observe applications without tying all instrumentation to a single provider.

Its name appears frequently when a website is divided into several services and discovering where a failure occurs starts to be difficult.

What problem does it solve?

Imagine that a purchase goes through an API, an inventory service, a payments service, and a database. For the user it is all just one button. For the technical team it is several different operations.

If the payment takes six seconds, you need to know where that time was lost. OpenTelemetry allows instrumenting those components consistently and relating information between them.

It does not automatically identify the cause of each error: it provides mechanisms to gather evidence.

What is a span?

A span represents an operation with a start and an end. It can correspond to an HTTP request, a database query, or a call between services.

Several related spans form a trace. The trace identifier allows correlating parts of the same operation, even if they cross different processes.

This requires propagating the context correctly. If a service loses those identifiers, part of the journey may appear disconnected.

Automatic and manual instrumentation

Some frameworks and libraries allow automatic instrumentation of common operations. This way you can get traces of HTTP requests without modifying each function.

But to measure a specific rule, such as "validate reservation availability," it may make sense to add manual instrumentation. A useful trace combines infrastructure data with operations that are meaningful to the business.

What is the Collector?

The ecosystem includes OpenTelemetry Collector, a component that can receive, process, and export telemetry signals.

It can be useful for centralizing the sending of information, applying transformations, or directing data to a compatible analysis platform. But it is another piece that needs to be configured, maintained, and supervised.

Do not confuse the Collector with a complete visualization interface: normally data is queried in a backend or specialized tool.

What about logs?

Logs help explain events. Their relationship with trace identifiers makes it easier to go from "the API returned an error" to "this specific operation failed within this service."

Not all libraries add context automatically and capabilities may vary by language and version. It is important to verify what instrumentation was actually implemented.

Costs, privacy, and sampling

Saving all traces of a highly concurrent application can consume storage and bandwidth. There are sampling strategies to keep only a portion, although that means you won't always have every request available.

Also, you shouldn't add passwords, tokens, or complete personal data as telemetry attributes. An observability tool can become a risk if it receives unnecessary secrets.

A useful example for diagnosing a failure

After updating an application, the latency metric increases. A trace shows that the backend takes little time to process the request, but an external call uses almost all the time.

Instead of buying a more powerful VPS on intuition, you can review the external provider, its response time, and the timeout policy.

How does it relate to ArduMaker?

If you host several services on VPSs managed from ArduMaker, you can instrument them with OpenTelemetry if their frameworks are compatible and send the signals to an observability tool. The main value lies in correlating behavior between applications, not in adding more charts for decoration.

In summary: OpenTelemetry standardizes part of the production and transport of telemetry. It is useful for investigating distributed systems, provided you instrument the correct operations and protect the collected data.

Sources: OpenTelemetry Documentation · Trace Concepts · Collector.