Use -f flag for targeted feature scraping to improve performance

[Mr-Bubz]

Problem

The exporter currently uses lmutil lmstat -a, which queries ALL features. On large license servers (e.g. Siemens with hundreds of features), this command is extremely slow and resource-intensive, often causing timeouts in Prometheus. A single scrape takes ~3 minutes to complete in my use case, even though features_to_include only lists 6 licenses.

Testing

Ran lmutil lmstat via command line.

• lmutil lmstat -a: 2 minutes
• lmutil lmstat -f feature_X: 10 seconds

Proposal

Modify the collector to utilize the features_to_include list from config.yml. If this list is present, the exporter should iterate through these features and execute lmutil lmstat -f instead of a global scrape. This shifts the filtering from post-processing (Regex) to the lmutil command itself.

I am willing to contribute a PR for this if the approach is approved.

[mjtrangoni]

Hi @Mr-Bubz,

Thanks for your input here! Indeed it seems very promising, and I am more than happy to support it.

I would like to avoid braking changes, as I think the config file is already very flexible, so my requirements are,

• Use the same config file format, if no features_to_include, and/or features_to_exclude, we should keep the -a behavior.
• New test cases with anonymized license files for every new case needs to be in place, and also be backwards compatible as well.
• Metrics exposed should stay the same.
• Check if there is a specific FlexLM version from which this -f is supported.

Thank you!