explain panic mode

Signed-off-by: Huabing Zhao <zhaohuabing@gmail.com>
zhaohuabing · Jan 2, 2025 · 0082053 · 0082053
1 parent f71fa99
commit 0082053
Show file tree

Hide file tree

Showing 3 changed files with 17 additions and 0 deletions.
diff --git a/api/v1alpha1/healthcheck_types.go b/api/v1alpha1/healthcheck_types.go
@@ -9,6 +9,11 @@ import metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 
 // HealthCheck configuration to decide which endpoints
 // are healthy and can be used for routing.
+//
+// Please note that Envoy load balancer may behave differently when lots of endpoints are unhealthy because of the "panic mode".
+// When the percentage of unhealthy endpoints exceeds 50%, Envoy will disregard health status and balance across all endpoints.
+// This is called "panic mode". It's designed to prevent a situation in which host failures cascade throughout the cluster
+// as load increases.
 type HealthCheck struct {
 	// Active health check configuration
 	// +optional

diff --git a/site/content/en/latest/api/extension_types.md b/site/content/en/latest/api/extension_types.md
@@ -2165,6 +2165,12 @@ _Appears in:_
 HealthCheck configuration to decide which endpoints
 are healthy and can be used for routing.
 
+
+Please note that Envoy load balancer may behave differently when lots of endpoints are unhealthy because of the "panic mode".
+When the percentage of unhealthy endpoints exceeds 50%, Envoy will disregard health status and balance across all endpoints.
+This is called "panic mode". It's designed to prevent a situation in which host failures cascade throughout the cluster
+as load increases.
+
 _Appears in:_
 - [BackendTrafficPolicySpec](#backendtrafficpolicyspec)
 - [ClusterSettings](#clustersettings)

diff --git a/site/content/zh/latest/api/extension_types.md b/site/content/zh/latest/api/extension_types.md
@@ -2165,6 +2165,12 @@ _Appears in:_
 HealthCheck configuration to decide which endpoints
 are healthy and can be used for routing.
 
+
+Please note that Envoy load balancer may behave differently when lots of endpoints are unhealthy because of the "panic mode".
+When the percentage of unhealthy endpoints exceeds 50%, Envoy will disregard health status and balance across all endpoints.
+This is called "panic mode". It's designed to prevent a situation in which host failures cascade throughout the cluster
+as load increases.
+
 _Appears in:_
 - [BackendTrafficPolicySpec](#backendtrafficpolicyspec)
 - [ClusterSettings](#clustersettings)