From 2a499b424f19af0ae7246f989c0a21f1773949ab Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=BC=A0=E7=A5=96=E5=BB=BA?= <zhangzujian.7@gmail.com>
Date: Fri, 15 Dec 2023 09:34:41 +0800
Subject: [PATCH] fix chassis gc (#3525)

Signed-off-by: zhangzujian <zhangzujian.7@gmail.com>
---
 pkg/controller/gc.go   | 12 +++++++++++-
 pkg/controller/node.go | 13 +++++++++++--
 2 files changed, 22 insertions(+), 3 deletions(-)

diff --git a/pkg/controller/gc.go b/pkg/controller/gc.go
index 19411036845..9476f3549b7 100644
--- a/pkg/controller/gc.go
+++ b/pkg/controller/gc.go
@@ -705,7 +705,7 @@ func (c *Controller) gcStaticRoute() error {
 
 func (c *Controller) gcChassis() error {
 	klog.Infof("start to gc chassis")
-	chassises, err := c.OVNSbClient.GetKubeOvnChassisses()
+	chassises, err := c.OVNSbClient.ListChassis()
 	if err != nil {
 		klog.Errorf("failed to get all chassis, %v", err)
 	}
@@ -727,6 +727,7 @@ func (c *Controller) gcChassis() error {
 		if hostname, exist := chassisNodes[chassisName]; exist {
 			if hostname == node.Name {
 				// node is alive, matched chassis should be alive
+				delete(chassisNodes, chassisName)
 				continue
 			}
 			// maybe node name changed, delete chassis
@@ -737,6 +738,15 @@ func (c *Controller) gcChassis() error {
 			}
 		}
 	}
+
+	for chassisName, hostname := range chassisNodes {
+		klog.Infof("gc node %s chassis %s", hostname, chassisName)
+		if err := c.OVNSbClient.DeleteChassis(chassisName); err != nil {
+			klog.Errorf("failed to delete node %s chassis %s %v", hostname, chassisName, err)
+			return err
+		}
+	}
+
 	return nil
 }
 
diff --git a/pkg/controller/node.go b/pkg/controller/node.go
index e171c984c60..c8880fc1985 100644
--- a/pkg/controller/node.go
+++ b/pkg/controller/node.go
@@ -992,17 +992,26 @@ func (c *Controller) UpdateChassisTag(node *v1.Node) error {
 		// kube-ovn-cni not ready to set chassis
 		return nil
 	}
-	chassis, err := c.OVNSbClient.GetChassis(annoChassisName, false)
+	chassis, err := c.OVNSbClient.GetChassis(annoChassisName, true)
 	if err != nil {
 		klog.Errorf("failed to get node %s chassis: %s, %v", node.Name, annoChassisName, err)
 		return err
 	}
+	if chassis == nil {
+		klog.Infof("chassis not registered for node %s, do chassis gc once", node.Name)
+		// chassis name conflict, do GC
+		if err = c.gcChassis(); err != nil {
+			klog.Errorf("failed to gc chassis: %v", err)
+			return err
+		}
+		return fmt.Errorf("chassis not registered for node %s, will try again later", node.Name)
+	}
+
 	if chassis.ExternalIDs == nil || chassis.ExternalIDs["vendor"] != util.CniTypeName {
 		klog.Infof("init tag %s for node %s chassis", util.CniTypeName, node.Name)
 		if err = c.OVNSbClient.UpdateChassisTag(chassis.Name, node.Name); err != nil {
 			return fmt.Errorf("failed to init chassis tag, %v", err)
 		}
-		return nil
 	}
 	return nil
 }