[xds] aggressive DNS refresh for faster tunnel endpoint propagation

dilyevsky · claude · dilyevsky · commit 5124f0f90952 · 2026-03-04T23:20:27.000-08:00
Lower STRICT_DNS cluster refresh from 30s to 2s, add failure refresh
rate (1-2s), and disable RespectDnsTtl so Envoy picks up new tunnel
endpoints quickly regardless of DNS TTL.

Co-Authored-By: Claude Opus 4.6 &lt;noreply@anthropic.com&gt;
diff --git a/pkg/gateway/xds/translator/cluster.go b/pkg/gateway/xds/translator/cluster.go
@@ -142,8 +142,12 @@ func buildXdsCluster(args *xdsClusterArgs) *clusterv3.Cluster {
 		}
 	} else {
 		cluster.ClusterDiscoveryType = &clusterv3.Cluster_Type{Type: clusterv3.Cluster_STRICT_DNS}
-		cluster.DnsRefreshRate = durationpb.New(30 * time.Second)
-		cluster.RespectDnsTtl = true
+		cluster.DnsRefreshRate = durationpb.New(2 * time.Second)
+		cluster.DnsFailureRefreshRate = &clusterv3.Cluster_RefreshRate{
+			BaseInterval: durationpb.New(1 * time.Second),
+			MaxInterval:  durationpb.New(2 * time.Second),
+		}
+		cluster.RespectDnsTtl = false
 	}
 
 	// build common, HTTP/1 and HTTP/2  protocol options for cluster

Original file line number	Diff line number	Diff line change
`@@ -142,8 +142,12 @@ func buildXdsCluster(args xdsClusterArgs) clusterv3.Cluster {`
`142`	`142`	`}`
`143`	`143`	`} else {`
`144`	`144`	`cluster.ClusterDiscoveryType = &clusterv3.Cluster_Type{Type: clusterv3.Cluster_STRICT_DNS}`
`145`		`- cluster.DnsRefreshRate = durationpb.New(30 * time.Second)`
`146`		`- cluster.RespectDnsTtl = true`
	`145`	`+ cluster.DnsRefreshRate = durationpb.New(2 * time.Second)`
	`146`	`+ cluster.DnsFailureRefreshRate = &clusterv3.Cluster_RefreshRate{`
	`147`	`+ BaseInterval: durationpb.New(1 * time.Second),`
	`148`	`+ MaxInterval: durationpb.New(2 * time.Second),`
	`149`	`+ }`
	`150`	`+ cluster.RespectDnsTtl = false`
`147`	`151`	`}`
`148`	`152`
`149`	`153`	`// build common, HTTP/1 and HTTP/2 protocol options for cluster`