GitPod Helm chart 0.15.0 raises workspace controller PostgreSQL max connections to 150

Gitpod Helm chart 0.15.0 does not expose a dedicated workspace-controller pool utilization metric by default. To obtain visibility, you must enable the optional metrics sidecar (typically via controller.metrics.enabled=true in your values file) and scrape the resulting Prometheus endpoint for pool-specific gauges.

Sizing GP_POSTGRES_MAX_CONNECTIONS

To determine the optimal value for GP_POSTGRES_MAX_CONNECTIONS, use a formula based on your peak concurrent workspace lifecycle events (starts and stops). Since each operation typically requires 1–2 active connections, the following calculation provides a baseline:

Limit = (Peak Concurrent Operations × 2) × 1.2 (Safety Margin)

Example: If your environment peaks at 50 concurrent workspace starts/stops, a limit of 120 connections (50 × 2 × 1.2) is recommended. Setting this value too high may lead to PostgreSQL memory exhaustion, while setting it too low will trigger PG::ConnectionBad errors during peak bursts.

Verifying Lazy Eviction

You can confirm that idle connections are being closed after five minutes without parsing controller logs by querying the PostgreSQL system catalogs directly. Run the following command against your database:

SELECT count(*) FROM pg_stat_activity WHERE state = 'idle' AND state_change < now() - interval '5 minutes';

Observe the result during a period of zero workspace activity. If lazy eviction is functioning as intended, the count of idle connections older than five minutes should trend toward zero as the controller releases them.

Operational Verification Steps

• Confirm Configuration: Verify the active limit using kubectl exec <workspace-controller-pod> -- env | grep GP_POSTGRES_MAX_CONNECTIONS.
• Monitor Saturation: Compare your peak workspace_start_total metrics against the configured limit to ensure utilization remains below 80%.

Missing Diagnostic: To refine this recommendation, please provide the average number of concurrent workspace operations observed during your highest peak load.