fix retries for node events in udn

See https://github.com/ovn-org/ovn-kubernetes/blob/master/go-controller/pkg/ovn/default_network_controller.go#L920
In UDN L3/L2 controllers we were missing
checking for mgmtPortFailed and accounting for that
as part of nodeupdates. So retry for any errors from
syncManagementPort was not happening.

Signed-off-by: Surya Seetharaman <suryaseetharaman.9@gmail.com>

go-controller/pkg/ovn/secondary_layer2_network_controller.go

@@ -159,16 +159,18 @@ func (h *secondaryLayer2NetworkControllerEventHandler) UpdateResource(oldObj, ne
 		if newNodeIsLocalZoneNode {
 			var nodeSyncsParam *nodeSyncs
 			if h.oc.isLocalZoneNode(oldNode) {
-				// determine what actually changed in this update
-				syncMgmtPort := macAddressChanged(oldNode, newNode, h.oc.NetInfo.GetNetworkName()) || nodeSubnetChanged
-
+				// determine what actually changed in this update and combine that with what failed previously
+				_, mgmtUpdateFailed := h.oc.mgmtPortFailed.Load(newNode.Name)
+				shouldSyncMgmtPort := mgmtUpdateFailed ||
+					macAddressChanged(oldNode, newNode, h.oc.NetInfo.GetNetworkName()) ||
+					nodeSubnetChanged
 				_, gwUpdateFailed := h.oc.gatewaysFailed.Load(newNode.Name)
 				shouldSyncGW := gwUpdateFailed ||
 					gatewayChanged(oldNode, newNode) ||
 					hostCIDRsChanged(oldNode, newNode) ||
 					nodeGatewayMTUSupportChanged(oldNode, newNode)
 
-				nodeSyncsParam = &nodeSyncs{syncMgmtPort: syncMgmtPort, syncGw: shouldSyncGW}
+				nodeSyncsParam = &nodeSyncs{syncMgmtPort: shouldSyncMgmtPort, syncGw: shouldSyncGW}
 			} else {
 				klog.Infof("Node %s moved from the remote zone %s to local zone %s.",
 					newNode.Name, util.GetNodeZone(oldNode), util.GetNodeZone(newNode))

go-controller/pkg/ovn/secondary_layer3_network_controller.go

@@ -168,6 +168,7 @@ func (h *secondaryLayer3NetworkControllerEventHandler) UpdateResource(oldObj, ne
 				_, nodeSync := h.oc.addNodeFailed.Load(newNode.Name)
 				_, failed := h.oc.nodeClusterRouterPortFailed.Load(newNode.Name)
 				clusterRtrSync := failed || nodeChassisChanged(oldNode, newNode) || nodeSubnetChanged
+				_, failed = h.oc.mgmtPortFailed.Load(newNode.Name)
 				syncMgmtPort := failed || macAddressChanged(oldNode, newNode, h.oc.GetNetworkName()) || nodeSubnetChanged
 				_, syncZoneIC := h.oc.syncZoneICFailed.Load(newNode.Name)
 				syncZoneIC = syncZoneIC || zoneClusterChanged