set nri always reconnect

Signed-off-by: Zhang Kang <kang.zhang@intel.com>
koordinator-sh · Apr 29, 2024 · 094d565 · 094d565
1 parent b46a168
commit 094d565
Show file tree

Hide file tree

Showing 3 changed files with 7 additions and 4 deletions.
diff --git a/pkg/koordlet/runtimehooks/config.go b/pkg/koordlet/runtimehooks/config.go
@@ -18,6 +18,7 @@ package runtimehooks
 
 import (
 	"flag"
+	"math"
 	"time"
 
 	"k8s.io/apimachinery/pkg/util/runtime"
@@ -134,8 +135,8 @@ func NewDefaultConfig() *Config {
 		RuntimeHooksNRI:                 true,
 		RuntimeHooksNRIConnectTimeout:   2 * time.Second,
 		RuntimeHooksNRIBackOffDuration:  1 * time.Second,
-		RuntimeHooksNRIBackOffCap:       10 * time.Minute,
-		RuntimeHooksNRIBackOffSteps:     100,
+		RuntimeHooksNRIBackOffCap:       1<<62 - 1,
+		RuntimeHooksNRIBackOffSteps:     math.MaxInt32,
 		RuntimeHooksNRIBackOffFactor:    2,
 		RuntimeHooksNRISocketPath:       "nri/nri.sock",
 		RuntimeHookReconcileInterval:    10 * time.Second,

diff --git a/pkg/koordlet/runtimehooks/config_test.go b/pkg/koordlet/runtimehooks/config_test.go
@@ -18,6 +18,7 @@ package runtimehooks
 
 import (
 	"flag"
+	"math"
 	"testing"
 	"time"
 
@@ -38,8 +39,8 @@ func Test_NewDefaultConfig(t *testing.T) {
 		RuntimeHooksNRI:                 true,
 		RuntimeHooksNRIConnectTimeout:   2 * time.Second,
 		RuntimeHooksNRIBackOffDuration:  1 * time.Second,
-		RuntimeHooksNRIBackOffCap:       10 * time.Minute,
-		RuntimeHooksNRIBackOffSteps:     100,
+		RuntimeHooksNRIBackOffCap:       1<<62 - 1,
+		RuntimeHooksNRIBackOffSteps:     math.MaxInt32,
 		RuntimeHooksNRIBackOffFactor:    2,
 		RuntimeHooksNRISocketPath:       "nri/nri.sock",
 		RuntimeHookReconcileInterval:    10 * time.Second,

diff --git a/pkg/koordlet/runtimehooks/nri/server.go b/pkg/koordlet/runtimehooks/nri/server.go
@@ -253,6 +253,7 @@ func (p *NriServer) onClose() {
 		}
 	}
 
+	// TODO: high version wait not support BackoffUntil with BackOffManger as parameters, when updated to v0.27.0 version wait, we can refine ExponentialBackoff.
 	err := wait.ExponentialBackoff(p.options.BackOff, retryFunc)
 	if err != nil {
 		klog.Errorf("nri server restart failed after several times retry")