Automated cherry pick of #4126: Add OFSwitch connection check to Agent's liveness probes

[tnqn]

Cherry pick of #4126 on release-1.7.

#4126: Add OFSwitch connection check to Agent's liveness probes

For details on the cherry pick process, see the cherry pick requests page.

[codecov]

Codecov Report

Merging #4447 (00d545e) into release-1.7 (693f568) will decrease coverage by 1.61%.
The diff coverage is 76.78%.

@@               Coverage Diff               @@
##           release-1.7    #4447      +/-   ##
===============================================
- Coverage        64.06%   62.44%   -1.62%     
===============================================
  Files              293      298       +5     
  Lines            43417    44428    +1011     
===============================================
- Hits             27815    27743      -72     
- Misses           13362    14464    +1102     
+ Partials          2240     2221      -19     

Flag	Coverage Δ
kind-e2e-tests	49.68% <86.20%> (-0.76%)	⬇️
unit-tests	44.14% <73.21%> (-0.54%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
cmd/antrea-agent/agent.go	0.00% <0.00%> (ø)
cmd/antrea-agent/options.go	4.50% <82.35%> (ø)
pkg/agent/apiserver/apiserver.go	71.27% <100.00%> (+4.25%)	⬆️
pkg/ovs/openflow/ofctrl_bridge.go	57.85% <100.00%> (+0.99%)	⬆️
pkg/agent/controller/networkpolicy/packetin.go	16.90% <0.00%> (-52.82%)	⬇️
pkg/agent/flowexporter/exporter/exporter.go	49.00% <0.00%> (-28.97%)	⬇️
pkg/agent/flowexporter/utils.go	44.68% <0.00%> (-27.66%)	⬇️
pkg/controller/networkpolicy/tier.go	50.00% <0.00%> (-5.00%)	⬇️
pkg/agent/util/ipset/ipset.go	67.56% <0.00%> (-2.71%)	⬇️
pkg/ovs/openflow/ofctrl_packetin.go	69.62% <0.00%> (-1.27%)	⬇️
... and 32 more

[wenyingd]

The unit test case "TestOFBridgeIsConnected" is failed, because ofnet is stuck at waiting for tlvMap reply in the versions earlier than this patch is merged antrea-io/ofnet#31 . To resolve the ut failure, we should modify some functions like this,

func newFakeOFSwitch(stream *util.MessageStream, app ofctrl.AppInterface) *ofctrl.OFSwitch {
	dpid, _ := net.ParseMAC("01:02:03:04:05:06:07:08")
	connCh := make(chan int)
	sw := ofctrl.NewSwitch(stream, dpid, app, connCh, 100)
	return sw
}

func sendTlvMapReply(stream *util.MessageStream) {
	reply := &openflow13.TLVTableReply{
		MaxSpace:  248,
		MaxFields: 62,
		TlvMaps: []*openflow13.TLVTableMap{
			{
				OptClass:  0xffff,
				OptType:   0,
				OptLength: 16,
				Index:     0,
			},
			{
				OptClass:  0xffff,
				OptType:   1,
				OptLength: 16,
				Index:     1,
			},
		},
	}
	tlvReplyMessage := openflow13.NewNXTVendorHeader(openflow13.Type_TlvTableReply)
	tlvReplyMessage.VendorData = reply
	stream.Inbound <- tlvReplyMessage
}

// TestOFBridgeIsConnected verifies it's thread-safe to call OFBridge's IsConnected method.
func TestOFBridgeIsConnected(t *testing.T) {
	stream := util.NewMessageStream(&fakeConn{}, nil)
	b := NewOFBridge("test-br", GetMgmtAddress(ovsconfig.DefaultOVSRunDir, "test-br"))
	var wg sync.WaitGroup
	wg.Add(1)
	go func() {
		defer wg.Done()
		newFakeOFSwitch(stream, b)
	}()
	go func() {
		select {
		case <-time.After(time.Millisecond * 100):
			sendTlvMapReply(stream)
		}
	}()
	wg.Add(1)
	go func() {
		defer wg.Done()
		b.IsConnected()
	}()
	wg.Wait()
}

OFBridge.Connected is called automatically in funciton ofctrl.NewSwitch, and no need to explicitly call it in the test.

[tnqn]

Thanks @wenyingd for the suggestion, will apply it.

[tnqn]

/test-latest-all

[xliuxu]

/test-latest-e2e

[tnqn]

e2e failed on #4477, which should be unrelated to the backport.