Add functionality to scrape kubelets metrics and capture npm metrics in case of network test

[agrawaliti]

This pull request introduces the ability to scrape kubelet metrics as part of the clusterloader2 SLO configuration and execution. The changes include updates to configuration files, Python scripts, and pipeline definitions to support this new functionality. Here's a breakdown of the most important changes:

Configuration Enhancements:

• Added a new parameter CL2_SCRAPE_KUBELETS in load-config.yaml to enable or disable scraping kubelet metrics. Conditional logic was added to include kubelet metrics modules during the "start" and "gather" phases. [1] [2] [3]

• Introduced a new module npm-measurements.yaml for Azure NPM metrics, which includes Prometheus queries for CPU and memory usage (average and maximum) at different percentiles (P99, P90, P50).

Python Script Updates:

• Updated the configure_clusterloader2 and execute_clusterloader2 functions in slo.py to handle the new scrape_kubelets parameter. This includes writing the parameter to the configuration file and passing it to the clusterloader2 execution command. [1] [2] [3]

• Modified the main() function in slo.py to accept scrape_kubelets as an argument for both the "configure" and "execute" commands, ensuring the parameter is properly parsed and passed. [1] [2] [3]

Pipeline and Step Updates:

• Updated the network-churn pipeline YAML to set scrape_kubelets to True for relevant stages.

• Adjusted the execute.yml step script to include the SCRAPE_KUBELETS environment variable when invoking the Python script for configuration and execution.

[agrawaliti]

cluster churn run: https://dev.azure.com/akstelescope/telescope/_build/results?buildId=23683&view=results
Service churn run: https://dev.azure.com/akstelescope/telescope/_build/results?buildId=23684&view=results