Preview: Server Metrics Dashboard

[mmcfarland]

We’re introducing a new Server Metrics Dashboard (preview) in the PostgreSQL extension for VS Code. You can now view server performance and health metrics without leaving your editor:

• All servers: real-time metrics from PostgreSQL system tables (calculated on the fly).
• Azure PostgreSQL Flexible Servers: additional historical metrics from Azure Monitor, at multiple time windows and resolutions.
• GitHub Copilot Chat: metrics are also exposed in Copilot, where our agent tools can provide initial diagnostic evaluations and performance analysis.

How to access it

Opt-in to the latest pre-release extension version, 1.9. See the full changelog here.

In the Object Explorer, right-click any connected server and select “Dashboard” from the context menu.

What’s next

This is our initial release. Upcoming improvements include:

• Additional metrics and support for customizing which metrics are shown.
• Integration with more diagnostic data sources, such as Azure Query Store and PostgreSQL’s pg_stat_statements.
• Enhanced diagnostic capabilities for our agent tools, enabling more advanced analysis of workload patterns, bottlenecks, and query performance using additional data sources.

We want your feedback

We’d especially like to hear:

• Which of the current metrics you find most useful.
• Any important gaps or missing metrics.
• How you’d prioritize future metrics and diagnostic capabilities.
• Chart customizations (specify aggregation method, filter by database, etc).

Please try it out in the latest pre-release build v1.9 and share your feedback in this discussion.

[ffrenkelzoo]

Good job, very usefull, thanks!

[mdresch]

I would be mainly interested in the combination of expensive queries and the underlaying performance of hardware. Whether the metrics would enable the scaling of available virtual hardware that can be scaled for high demanding but rare jobs that only need to be performed occasionally. The relation between the connections, CPU and Memory usage to scale out the virtual underlaying hardware to enable a scalable infrastructure to automate these tasks. Maybe this will enable the system to go from cold start upon login to modular requests from applications and scale hardware requirements inline with the modules or expensive queries that need to run through the systems.