Inconsistency in identifying labels

[opslogixdennis]

Problem Statement

My original question was specifically for Grafana Alloy, but I was suggested to post my issue here.

I've been tackling a problem regarding the inconsistency in labels present to identify a host. The prometheus.exporter.windows component only adds identifiers to the metrics windows_cs_hostname and windows_os_hostname, and does not add any identifier at all to other metrics.

One might add the fqdn statically directly to the relabel job, or by adding it as a environment variable accessible by e.g. sys.getenv("FQDN").

Would it be possible to add a label dynamically?

Proposed Solution

No response

Additional information

No response

Acceptance Criteria

No response

[jkroepke]

Prometheus or Alloy should add an instance label to all metrics.

[opslogixdennis]

Prometheus or Alloy should add an instance label to all metrics.

Yes, but that is only the hostname. We would like the host domain name, either directly in instance as fqdn or as a separate label. The metric windows_cs_hostname contains this information.

All other instance labels only use the hostname, not fqdn.

EDIT:

The promtheus node exporter adds fqdn as instance label.

[jkroepke]

The promtheus node exporter adds fqdn as instance label.

promtheus node exporter does not expose any instance labels. It comes, because you are using fqdn as target in your prometheus or ally config.

In that context, you can do this:

windows_metric * on(instance) group_left(domain, fqdn) windows_os_hostname

Then, you have the labels available on any metric you wish.

See: https://stackoverflow.com/questions/44461801/how-can-i-join-two-metrics-in-a-prometheus-query

[opslogixdennis]

The promtheus node exporter adds fqdn as instance label.

promtheus node exporter does not expose any instance labels. It comes, because you are using fqdn as target in your prometheus or ally config.

In that context, you can do this:

windows_metric * on(instance) group_left(domain, fqdn) windows_os_hostname

Then, you have the labels available on any metric you wish.

See: https://stackoverflow.com/questions/44461801/how-can-i-join-two-metrics-in-a-prometheus-query

My bad, you are correct. I was looking at the wrong config file.

The solution seem to work and solves the issue. Maybe it could be added to the main documentation?

Thank you very much for the help!