What is Grafana and how to create dashboards that are useful for investigating problems?

Published on 2026-10-21 by frfrand

Grafana explicado: fuentes de datos, dashboards, alertas y errores al visualizar métricas sin contexto.

Grafana is a platform for querying and visualizing observability data through dashboards, charts, and alerts. It is frequently used to monitor servers, databases, and applications, but it does not work as a universal data source by itself: it normally needs to connect to systems that already store the information.

Its usefulness begins when you stop looking at a chart because it looks pretty and start asking questions that really need answers.

The example of a control panel

Imagine a transportation company. It has fuel, delay, and maintenance information in different systems. If each person has to log into ten programs to understand what is happening, investigating a problem takes too long.

Grafana allows you to build views that gather data queried from compatible sources. One panel can show API latency and another the memory consumption of servers.

A dashboard is not a copy of all the data. It is a representation of queries on information sources.

What is a data source?

A data source is the configured connection to a system that stores data. Grafana can work with Prometheus, PostgreSQL, Loki, cloud systems, and other sources via compatible plugins.

This means that installing Grafana and waiting for it to discover all your errors is not enough. First you need to collect information and configure permissions to query the correct sources.

The data plugin determines how the query is written and what results it returns.

How to build a useful dashboard?

Start with a question:

"Did errors increase after the latest deployment?"

To answer it, you might need a chart of failed requests per minute, a comparison with total requests, and an annotation marking when the new version was published.

That dashboard is more useful than a screen full of gauges showing numbers without context.

Another example: an online store observes its API response time. An average can hide that some shoppers wait ten seconds. Therefore, depending on available metrics, it may be useful to show latency percentiles.

What are variables and filters?

A dashboard can allow selecting the environment —development or production—, the instance, or the service. This avoids creating dozens of nearly identical panels.

But labels need a coherent structure. If one system identifies the environment as prod and another as production, filters can become confusing.

Defining consistent names is a small task that greatly improves investigation.

Can Grafana send alerts?

Yes, it includes alerting capabilities that work with data sources and configured rules. You can define conditions, notification routes, and other controls according to the deployment.

The difficulty is not creating an alert that says "High CPU". It is deciding when that condition requires intervention and how to avoid repeated notifications without value.

If no one knows what to do when an alert arrives, perhaps you need to document the procedure before adding more rules.

How to avoid misleading graphs?

Check units, intervals, and aggregations. A graph in milliseconds can seem alarming if the person thinks it shows seconds. Comparing a five-minute average with an instantaneous spike is not always reasonable either.

Define relevant periods and mark version changes. It is also advisable to review missing data: an empty line does not prove that the system had zero errors.

What about security?

A dashboard can expose client metrics, internal names, and operational details. Control who can query each source and who can share panels.

Do not assume that hiding a folder in the interface is enough if the data source permissions allow querying information from other companies.

And in ArduMaker?

In applications hosted on ArduMaker VPS, Grafana can be used as a visualization tool for metrics collected by compatible services. The hosting platform does not decide which panels your business needs: first you must define questions, metrics, and permissions.

In one sentence: Grafana transforms queries into understandable views. A good dashboard helps make a decision, not just decorate a screen.

Official sources: Data sources in Grafana · Introduction to dashboards.