What does the job label mean in Prometheus?

The job label normally comes from job_name in a Prometheus scrape configuration. It identifies a group of targets collected under that configuration, such as Node Exporter instances or an application's metrics endpoints.

Job identifies the group; instance identifies the target

scrape_configs:
  - job_name: node
    static_configs:
      - targets: ['10.0.0.1:9100', '10.0.0.2:9100']

The targets normally produce samples labeled job="node", with instance="10.0.0.1:9100" or instance="10.0.0.2:9100" respectively.

Use the job label to select that group:

up{job="node"}

For an application counter with consistent meaning across its jobs, compare rates by job:

topk(5, sum by (job) (rate(http_requests_total[5m])))

This ranks request rates, not a generic measure of server load. Choose a metric whose semantics match the comparison.

Choose stable, descriptive names

Names such as mysql, node, or api_gateway make queries readable. A consistent lowercase convention is useful, but underscores are a team convention rather than a requirement for job label values.

Use additional labels such as environment="production" when you want to compare the same component across environments. Environment-specific job names can also work if the team uses them consistently. Avoid mixing unrelated purposes in one job when they require different scrape settings or make queries ambiguous.

Understand overrides and conflicts

Target relabeling can rewrite job, so the final label is not guaranteed to equal the original job_name.

If an exporter supplies its own job label, the default honor_labels: false keeps the target-side value and renames the conflicting scraped label with an exported_ prefix. Setting honor_labels: true preserves the exporter's conflicting label instead. This matters for federation and Pushgateway use cases.

FAQ

Can I omit job_name? A scrape configuration requires a job_name, and job names must be unique within the configuration. That configuration requirement is distinct from the final sample labels, which relabeling can transform.

Can one job scrape different ports? Yes. Each target can specify its own port. Split jobs when the endpoints need different scrape behavior or represent unrelated kinds of data, rather than solely because their port numbers differ.

Sources

This guide is based on the documentation below; the examples have not been tested against a live deployment.

• Prometheus jobs and instances
• Prometheus scrape configuration