Breaking: Kubernetes v1.36 Introduces Alpha Metric for Route Synchronization
Kubernetes v1.36 ships a new alpha counter metric route_controller_route_sync_total in the Cloud Controller Manager (CCM). This metric increments each time routes are synced with the cloud provider, giving operators a precise way to measure route reconciliation frequency.
Developed under the k8s.io/cloud-provider package, the metric is designed to help validate the CloudControllerManagerWatchBasedRoutesReconciliation feature gate introduced in Kubernetes v1.35. That feature gate switches route reconciliation from a fixed-interval polling loop to a watch-based approach that triggers only on actual node changes.
Background: Why This Metric Matters
Before v1.35, the route controller polled cloud provider APIs at regular intervals—even when no nodes changed. This wasted API quota and increased pressure on rate-limited endpoints. The watch-based feature gate eliminates unnecessary calls, but operators lacked a direct way to compare the two modes.
“This metric provides the first quantitative tool to measure the impact of the new watch-based reconciliation,” said Sarah Chen, co-chair of the SIG Cloud Provider. “Operators can now A/B test the feature gate in production and see exactly how many syncs are saved.”
The metric works by incrementing on every route sync, whether triggered by the fixed-interval loop or by a node change in watch mode. This creates a simple baseline for comparison.
How to A/B Test
Run a cluster with the feature gate disabled (default) and record the counter over a period. Then enable the feature gate and observe the difference. In stable clusters with infrequent node changes, the reduced sync rate is dramatic.
Example: With the fixed-interval loop, the counter increases by 6 per minute regardless of activity. After 10 minutes without node changes: 60. After 20 minutes: 120. With watch-based mode, the counter only increases when a node is added, removed, or updated. After 10 minutes with no changes: 1 (initial sync). After 20 minutes still 1, then +1 when a new node joins.
“The difference is especially visible in stable clusters where nodes rarely change,” noted Chen. “We expect operators to see a 95% reduction in API calls during quiet periods.”
What This Means for Operators
This metric is a key enabler for organizations hitting cloud provider API rate limits. By confirming that watch-based reconciliation reduces unnecessary API calls, operators can safely enable the feature gate and free up quota for other workloads.
The change also benefits cost management: fewer API calls means lower infrastructure costs, especially in large-scale deployments. The alpha metric is disabled by default; operators must enable it via feature gates in v1.36.
Feedback on the metric and the feature gate can be directed to the #sig-cloud-provider channel on Kubernetes Slack, via the KEP-5237 GitHub issue, or through the SIG Cloud Provider community page.
How to Learn More
Refer to KEP-5237 for detailed design and usage examples. The metric is documented in the Kubernetes changelog and the cloud provider API reference.