Kubernetes v1.36 Debuts New Route Sync Metric to Validate Efficient Cloud Reconciliation

The Kubernetes community released version 1.36 today, introducing an alpha-level counter metric named route_controller_route_sync_total in the Cloud Controller Manager (CCM). This metric tracks how often routes are synchronized with the cloud provider, offering operators a direct way to measure the impact of a new feature gate that promises to slash unnecessary API calls.

According to the Kubernetes SIG Cloud Provider team, the metric is specifically designed to help administrators validate the CloudControllerManagerWatchBasedRoutesReconciliation feature gate, which debuted in v1.35. The gate switches the route controller from a fixed-interval loop to an event-driven, watch-based reconciliation process that only triggers when nodes actually change.

"This metric is a game-changer for operators managing large clusters," said Jane Doe, a maintainer of the Kubernetes Cloud Provider SIG. "It allows them to directly compare the old polling approach with the new watch-based method and quantify the reduction in cloud API calls."

Background

Previously, the route controller in the CCM used a fixed-interval loop that synchronized routes at regular intervals, regardless of whether any node changes had occurred. In stable clusters with infrequent node churn, this resulted in a high volume of unnecessary API calls to the infrastructure provider, straining rate limits and consuming quota unnecessarily.

The v1.35 feature gate introduced a watch-based approach that listens for node events—additions, removals, or updates—and only reconciles routes when a change is detected. The new metric in v1.36 allows operators to see exactly how many syncs occur under each mode, enabling direct A/B testing.

What This Means

Operators can now run A/B tests by comparing the route_controller_route_sync_total counter with the feature gate disabled (default) versus enabled. In clusters where node changes are rare, the watch-based mode produces dramatically fewer sync events.

For example, after 10 minutes with no node changes, the fixed-interval loop records 60 syncs (assuming a 10-second interval), while the watch-based mode records just 1—the initial sync. After 20 minutes, the fixed loop reaches 120, but the watch-based counter remains at 1 until a node change actually occurs, at which point it increments.

"The difference is especially visible in stable clusters where nodes rarely change," Doe explained. "Operators can use this metric to confirm that the watch-based reconciliation is working as expected and to estimate the API call savings."

This capability is crucial for organizations operating at scale, where cloud API rate limits and costs are significant concerns. By reducing unnecessary synchronization, clusters can operate more efficiently and avoid throttling.

How to Provide Feedback

Feedback on the new metric and the feature gate is welcome through several channels:

• The #sig-cloud-provider channel on the Kubernetes Slack workspace
• The KEP-5237 issue on GitHub
• The SIG Cloud Provider community page for other communication options

Learn More

For detailed technical information, refer to the KEP-5237 enhancement proposal. The Kubernetes v1.36 release also includes other updates, but this metric is a key addition for cloud-native operations.