[feat/1316] Support Serving Metrics On an Alternate Port

[divyabiyani]

Feature Request: Jupyter Server Issue #1316 — Support Serving Metrics On an Alternate Port

Label suggestion: enhancement

Summary Of Changes:

• Introduced support for serving metrics on an alternate port.
• Updated the default behavior to publish metrics on port 9090.
• The port can be customized using the --ServerApp.metrics_port=<port> argument. Setting --ServerApp.metrics_port=0 will disable the separate server and serve metrics through the main Tornado server.
• The metrics server honors the following configuration flags:
  • --ServerApp.authenticate_prometheus=<bool> — Enables authentication for metrics endpoint if set to true.
  • --ServerApp.record_http_request_metrics=<bool> — Enables or disables recording of HTTP request metrics.
• Enhanced info logs to include the metrics URL for better visibility.

Test Scenarios Verified:

• Info logs correctly include the metrics URL.

• Verified that when a custom metrics port is provided, metrics are only served on the specified port and not on the default port or tornado server port.

• Verified behavior when no ServerApp.token='' or --ServerApp.password='' is provided:

  • With --ServerApp.authenticate_prometheus=False.

• With --ServerApp.authenticate_prometheus=True.

• Verified behavior when both ServerApp.token and ServerApp.password are provided.
• Verified functionality with --ServerApp.record_http_request_metrics=True.

• Verified metrics served with --ServerApp.metrics_port=9009.

• Verified fallback behavior with --ServerApp.metrics_port=0 (metrics served on main Tornado server).

• Verified that when the configured metrics port is already in use, server startup fails with appropriate error.