I have stability problems with Spring web-sockets. Sometimes I have a similar exception to that describe in https://jira.spring.io/browse/SPR-12812.
The patch is not available, then I have implemented my own code with a custom SubProtocolWebSocketHandler.
public class WBSSubProtocolWebSocketHandler extends SubProtocolWebSocketHandler {
private static final Logger LOG = LoggerFactory.getLogger(WBSSubProtocolWebSocketHandler.class);
public WBSSubProtocolWebSocketHandler(MessageChannel clientInboundChannel, SubscribableChannel clientOutboundChannel) {
super(clientInboundChannel, clientOutboundChannel);
}
#Override
public void afterConnectionClosed(WebSocketSession session, CloseStatus status) throws Exception {
super.afterConnectionClosed(session, status);
LOG.debug("WebSocket Connection closed for client with session ID {}", session.getId());
}
#Override
public void afterConnectionEstablished(WebSocketSession session) throws Exception {
// WebSocketHandlerDecorator could close the session
// https://jira.spring.io/browse/SPR-12812
if (!session.isOpen()) {
LOG.warn("WebSocket Connection established for client with session ID {} was closed.", session.getId());
return;
}
super.afterConnectionEstablished(session);
LOG.debug("WebSocket Connection established for client with session ID {}", session.getId());
}
#Override
public void handleTransportError(WebSocketSession session, Throwable exception) throws Exception {
super.handleTransportError(session, exception);
LOG.warn("WebSocket transport error for client with session ID {}", session.getId());
}
#Override
public void handleMessage(WebSocketSession session, WebSocketMessage message) throws Exception {
super.handleMessage(session, message);
LOG.debug("Websocket incoming message ({}) from client with session ID {}", message.getPayload().toString(), session.getId());
}
#Override
public void handleMessage(Message message) throws MessagingException {
super.handleMessage(message);
LOG.debug("Websocket incoming message : {}, header {}", message);
}
}
Now my problem is to reproduce the exception to see if the problem is solved. I tried various ways, but without success. I can not reproduce the closing of the connection. Does anyone have an idea?
We also have a second problem. The client application (angularjs application) sometimes reports that the socket-web connection is lost. But I do not understand why because on the server I have no error / warning in the logs.
How can I identify the problem and reproduce it?
Related
I'm making multi rooms chat with user authorization: users can have access only to some assigned rooms.
For every room I creating a topic with unique room id
How can I check permissions during the opening socket for reading?
On the server-side, for new inbound connection, I want to get room id from topic URL and check user access permissions for the room. But I didn't find how I can do it. I don't see the place, there it's possible.
AbstractSecurityWebSocketMessageBrokerConfigurer -- no way for dynamic check
#Configuration
class WebSocketSecurityConfig extends AbstractSecurityWebSocketMessageBrokerConfigurer {
#Override
protected void configureInbound(MessageSecurityMetadataSourceRegistry message) {
message.nullDestMatcher().permitAll()
.simpDestMatchers("/app/**").authenticated()
.anyMessage().hasRole("USER")
}
}
WebSocketMessageBrokerConfigurer -- can't get current url
#Configuration
class WebSocketSecurityConfig extends WebSocketMessageBrokerConfigurer {
#Override
public void configureClientInboundChannel(ChannelRegistration registration) {
registration.interceptors(new ChannelInterceptor() {
#Override
public Message<?> preSend(Message<?> message, MessageChannel channel) {
StompHeaderAccessor accessor = MessageHeaderAccessor.getAccessor(message, StompHeaderAccessor.class);
if (StompCommand.SUBSCRIBE.equals(accessor.getCommand())) {
...
}
return message
}
});
}
}
I know, how to check access during writing messages, but can't find, how to do it during opening a web socket for reading. What is the standard mechanism for this case?
Dependencies:
compile 'org.grails.plugins:grails-spring-websocket:2.5.0.RC1'
compile "org.springframework.security:spring-security-messaging"
compile "org.springframework.security:spring-security-config"
compile "org.springframework.security:spring-security-core:5.1.8.RELEASE"
compile "org.springframework:spring-messaging:5.1.6.RELEASE"
UPDATE
I can pass room id from the client as a header, but on the server in configureClientInboundChannel I can't be sure, that room id in header same with id in topic URL. I can use some hashes, generated on the server-side, but it looks too complex
var socket = new SockJS("${createLink(uri: '/stomp')}");
var client = webstomp.over(socket);
client.connect({room-id:"0"}, function() {
client.subscribe("/topic/room/1", function(message) {
console.log("/topic/room/1");
}, {roomId:"1"});
client.subscribe("/topic/room/2", function(message) {
console.log("/topic/room/2");
}, {roomId:"2"});
});
During debugging, I have checked headers of command with type StompCommand.CONNECT.
For StompCommand.SUBSCRIBE command current topic URL presented in simpDestination header
Final solution is:
#Configuration
class WebSocketSecurityConfig extends WebSocketMessageBrokerConfigurer {
#Override
public void configureClientInboundChannel(ChannelRegistration registration) {
registration.interceptors(new ChannelInterceptor() {
#Override
public Message<?> preSend(Message<?> message, MessageChannel channel) {
StompHeaderAccessor accessor = MessageHeaderAccessor.getAccessor(message, StompHeaderAccessor.class);
if (StompCommand.SUBSCRIBE.equals(accessor.getCommand())) {
def currentAuthentication = accessor.getHeader('simpUser') // from spring security
String destinationUrl = (String )accessor.getHeader('simpDestination')
// do check, and throw AuthenticationException
}
return message
}
});
}
}
I am trying to connect to a Docker UNIX domain socket using Netty. Here's my attempt so far.
#PostConstruct
public void init() throws Exception {
io.netty.bootstrap.Bootstrap bootstrap = new io.netty.bootstrap.Bootstrap();
bootstrap
.group(new NioEventLoopGroup())
.channel(NioSocketChannel.class)
.option(ChannelOption.SO_KEEPALIVE, true)
.remoteAddress(new DomainSocketAddress("/var/run/docker.sock"))
.handler(new ChannelInitializer<SocketChannel>() {
#Override
protected void initChannel(SocketChannel socketChannel) throws Exception {
socketChannel
.pipeline()
.addLast(new SimpleChannelInboundHandler<HttpObject>() {
#Override
protected void channelRead0(ChannelHandlerContext channelHandlerContext, HttpObject httpObject) throws Exception {
System.out.println(httpObject);
}
});
}
});
final Channel channel = bootstrap.connect().sync().channel();
final FullHttpRequest request = new DefaultFullHttpRequest(HttpVersion.HTTP_1_1, HttpMethod.GET, "/services", Unpooled.EMPTY_BUFFER);
request.headers().set(HttpHeaderNames.CONNECTION, HttpHeaderValues.CLOSE);
channel.writeAndFlush(request);
channel.closeFuture().sync();
System.out.println("DONE");
}
At the moment I am getting
Caused by: java.nio.channels.UnsupportedAddressTypeException: null
Is there an example on how to do HTTP connections to UDS using Netty? So far I only found raw UDS and TCP HTTP but not combined.
Here's a working implementation.
io.netty.bootstrap.Bootstrap bootstrap = new io.netty.bootstrap.Bootstrap();
final EpollEventLoopGroup epollEventLoopGroup = new EpollEventLoopGroup();
try {
bootstrap
.group(epollEventLoopGroup)
.channel(EpollDomainSocketChannel.class)
.handler(new ChannelInitializer<UnixChannel>() {
#Override
public void initChannel(UnixChannel ch) throws Exception {
ch
.pipeline()
.addLast(new HttpClientCodec())
.addLast(new HttpContentDecompressor())
.addLast(new SimpleChannelInboundHandler<HttpObject>() {
private StringBuilder messageBuilder = new StringBuilder();
#Override
public void channelRead0(ChannelHandlerContext ctx, HttpObject msg) throws Exception {
if (msg instanceof HttpContent) {
HttpContent content = (HttpContent) msg;
messageBuilder.append(content.content().toString(StandardCharsets.UTF_8));
if (msg instanceof LastHttpContent) {
System.out.println(messageBuilder);
}
} else {
System.out.println(msg.getClass());
}
}
});
}
});
final Channel channel = bootstrap.connect(new DomainSocketAddress("/var/run/docker.sock")).sync().channel();
final FullHttpRequest request = new DefaultFullHttpRequest(HttpVersion.HTTP_1_1, HttpMethod.GET, "/services", Unpooled.EMPTY_BUFFER);
request.headers().set(HttpHeaderNames.HOST, "daemon");
channel.writeAndFlush(request);
channel.closeFuture().sync();
} finally {
epollEventLoopGroup.shutdownGracefully();
}
Few things to note:
Use the EpollEventLoopGroup and EpollDomainSocketChannel with a ChannelInitializer<UnixChannel>.
HTTP requires new HttpCodec() in the pipeline to use the Netty HTTP objects.
The data may be chunked so you need to assemble it and wait for the LastHttpContent object
https://github.com/trajano/netty-docker-daemon-socket
i am not sure my understanding to errorHandler and returnExceptions is right or not.
but here is my goal: i sent a message from App_A, use #RabbitListener to receive message in App_B.
according to the doc
https://docs.spring.io/spring-amqp/docs/2.1.3.BUILD-SNAPSHOT/reference/html/_reference.html#annotation-error-handling
i assume if APP_B has a business exception during process the message,through set errorHandler and returnExceptions in a right way on #RabbitListener can let the exception back to App_A.
do I understood correctly?
if i am rigth, how to use it in a right way?
with my code, i get nothing in APP_A .
here is my code in APP_B
errorHandler:
#Component(value = "errorHandler")
public class ErrorHandler implements RabbitListenerErrorHandler {
#Override
public Object handleError(Message arg0, org.springframework.messaging.Message<?> arg1,
ListenerExecutionFailedException arg2) throws ListenerExecutionFailedException {
throw new ListenerExecutionFailedException("msg", arg2, null);
}
}
RabbitListener:
#RabbitListener(
bindings = #QueueBinding(
value = #Queue(value = "MRO.updateBaseInfo.queue", durable = "true"),
exchange = #Exchange(name = "MRO_Exchange", type = ExchangeTypes.DIRECT, durable = "true"),
key = "baseInfoUpdate"
),
// errorHandler = "errorHandler",
returnExceptions = "true"
)
public void receiveLocationChangeMessage(String message){
BaseUpdateMessage newBaseInfo = JSON.parseObject(message, BaseUpdateMessage.class);
dao.upDateBaseInfo(newBaseInfo);
}
and code in APP_A
#Component
public class MessageSender {
#Autowired
private RabbitTemplate rabbitTemplate;
public void editBaseInfo(BaseUpdateMessage message)throws Exception {
//and i am not sure set RemoteInvocationAwareMessageConverterAdapter in this way is right
rabbitTemplate.setMessageConverter(new RemoteInvocationAwareMessageConverterAdapter());
rabbitTemplate.convertAndSend("MRO_Exchange", "baseInfoUpdate", JSON.toJSONString(message));
}
}
i am very confuse with three points:
1)do i have to use errorHandler and returnExceptions at the same time? i thought errorHandler is something like a postprocessor that let me custom exception.if i don't need a custom exception can i just set returnExceptions with out errorHandler ?
2)should the method annotated with #RabbitListener return something or void is just fine?
3)in the sender side(my situation is APP_A), does have any specific config to catch the exception?
my workspace environment:
Spring boot 2.1.0
rabbitMQ server 3.7.8 on docker
1) No, you don't need en error handler, unless you want to enhance the exception.
2) If the method returns void; the sender will end up waiting for timeout for a reply that will never arrive, just in case an exception might be thrown; that is probably not a good use of resources. It's better to always send a reply, to free up the publisher side.
3) Just the RemoteInvocationAwareMessageConverterAdapter.
Here's an example:
#SpringBootApplication
public class So53846303Application {
public static void main(String[] args) {
SpringApplication.run(So53846303Application.class, args);
}
#RabbitListener(queues = "foo", returnExceptions = "true")
public String listen(String in) {
throw new RuntimeException("foo");
}
#Bean
public ApplicationRunner runner(RabbitTemplate template) {
template.setMessageConverter(new RemoteInvocationAwareMessageConverterAdapter());
return args -> {
try {
template.convertSendAndReceive("foo", "bar");
}
catch (Exception e) {
e.printStackTrace();
}
};
}
}
and
org.springframework.amqp.AmqpRemoteException: java.lang.RuntimeException: foo
at org.springframework.amqp.support.converter.RemoteInvocationAwareMessageConverterAdapter.fromMessage(RemoteInvocationAwareMessageConverterAdapter.java:74)
at org.springframework.amqp.rabbit.core.RabbitTemplate.convertSendAndReceive(RabbitTemplate.java:1500)
at org.springframework.amqp.rabbit.core.RabbitTemplate.convertSendAndReceive(RabbitTemplate.java:1433)
at org.springframework.amqp.rabbit.core.RabbitTemplate.convertSendAndReceive(RabbitTemplate.java:1425)
at com.example.So53846303Application.lambda$0(So53846303Application.java:28)
at org.springframework.boot.SpringApplication.callRunner(SpringApplication.java:804)
at org.springframework.boot.SpringApplication.callRunners(SpringApplication.java:794)
at org.springframework.boot.SpringApplication.run(SpringApplication.java:324)
at org.springframework.boot.SpringApplication.run(SpringApplication.java:1260)
at org.springframework.boot.SpringApplication.run(SpringApplication.java:1248)
at com.example.So53846303Application.main(So53846303Application.java:15)
Caused by: java.lang.RuntimeException: foo
at com.example.So53846303Application.listen(So53846303Application.java:20)
As you can see, there is a local org.springframework.amqp.AmqpRemoteException with the cause being the actual exception thrown on the remote server.
I have a Broadcast receiver that checks WIFI_STATE_CHANGE to see if I have connected to a certain WiFi network. For example if I am coming home, I want a certain MQTT message to be sent. The problem I have is that it connects and sends the MQTT message, only when run the app the first time.
Process:
If I build the application and run it on the device and it recognised my home WiFi it sends the message.
I turn off Wifi from the device, and turn it back on again.
I get "Failure" which is a message when the MQTT connection to the server could not be established.
What I would need is that after I reconnect to the network, instead of "Failure" to get "Connected" but somehow it never happens...what could be wrong?
PS. I think it has to do with the fact that when WiFi is detected, the Broadcast Receiver runs the connection code, although Internet is not available at that point of time (obtaining IP etc.)
Here is the code of the Broadcast receiver:
package me.app.comehomedemo;
import ...
public class SynchronizeBroadcastReceiver extends BroadcastReceiver {
MqttAndroidClient client;
static String MQTTHOST = "myhost";
static String USERNAME = "myusername";
static String PASSWORD = "mypassword";
static String topicStr = "/topic/mac/control";
static String payload = "1";
#Override
public void onReceive(final Context context, Intent intent) {
NetworkInfo info = intent.getParcelableExtra(WifiManager.EXTRA_NETWORK_INFO);
if (info.isConnected()) {
WifiManager wifiManager = ( WifiManager ) context.getSystemService(Context.WIFI_SERVICE);
WifiInfo wifiInfo = wifiManager.getConnectionInfo();
int ip = wifiInfo.getIpAddress();
Toast.makeText(context, String.valueOf(ip), Toast.LENGTH_SHORT).show();
String ssid = wifiInfo.getSSID();
if (ssid.equals("\"mySSID\"")) {
String clientId = MqttClient.generateClientId();
client = new MqttAndroidClient(context.getApplicationContext(), MQTTHOST, clientId);
MqttConnectOptions options = new MqttConnectOptions();
options.setUserName(USERNAME);
options.setPassword(PASSWORD.toCharArray());
// options.setAutomaticReconnect(true);
try {
IMqttToken token = client.connect(options);
token.setActionCallback(new IMqttActionListener() {
#Override
public void onSuccess(IMqttToken asyncActionToken) {
// We are connected
Toast.makeText(context, "Connected", Toast.LENGTH_SHORT).show();
try {
client.publish(topicStr, payload.getBytes(), 0, false);
} catch (MqttException e) {
e.printStackTrace();
}
}
#Override
public void onFailure(IMqttToken asyncActionToken, Throwable exception) {
// Something went wrong e.g. connection timeout or firewall problems
Toast.makeText(context, "Failure", Toast.LENGTH_SHORT).show();
}
});
} catch (MqttException e) {
e.printStackTrace();
}
Uri notification = RingtoneManager.getDefaultUri(RingtoneManager.TYPE_NOTIFICATION);
MediaPlayer mp = MediaPlayer.create(context.getApplicationContext(), notification);
mp.start();
}
}
}
}
I have managed to solve it by waiting 2 seconds and then running the task. Used this solution and it worked. I had to wait for the Internet connection to get ready!
Since waiting 2 seconds has solved your problem, then it might be that the Wifi broadcast comes too early, before there is a connection established (like DHCP gives your phone IP and establishes the routes) for the MQTT connect and publish packets to be properly delivered.
But what happens if some other user needs to wait 10 and not 2 seconds?
My suggestion is to set the automatic reconnect option in MqttConnectOptions and then use the connection callback to publish the needed info to the broker and finally disconnect in publish callback:
private IMqttActionListener mConnectCallback = new IMqttActionListener() {
#Override
public void onSuccess(IMqttToken token) {
try {
client.publish(topicStr, new MqttMessage(payload.getBytes()), null, mPublishCallback);
} catch (Exception ex) {
ex.printStackTrace();
}
}
#Override
public void onFailure(IMqttToken token, Throwable ex) {
}
};
private IMqttActionListener mPublishCallback = new IMqttActionListener() {
#Override
public void onSuccess(IMqttToken token) {
// TODO disconnect
}
#Override
public void onFailure(IMqttToken token, Throwable ex) {
}
};
MqttAndroidClient client = new MqttAndroidClient(context, MQTTHOST, "my_id");
MqttConnectOptions options = new MqttConnectOptions();
options.setUserName(USERNAME);
options.setPassword(PASSWORD.toCharArray());
options.setAutomaticReconnect(true);
client.connect(options, null, mConnectCallback);
Even though this is not described in the Spring documentation, a websocket connect should lead to a connection upgrade response (101 status).
#Configuration
#EnableWebSocketMessageBroker
public class WebSocketConfig<S extends ExpiringSession> extends AbstractSessionWebSocketMessageBrokerConfigurer<S>{
#Override
public void configureMessageBroker(MessageBrokerRegistry config) {
config.enableSimpleBroker("/topic", "/queue");
config.setApplicationDestinationPrefixes("/mobile-server");
config.setUserDestinationPrefix("/mobile-user");
}
#Override
public void configureStompEndpoints(StompEndpointRegistry registry) {
registry
.addEndpoint("/ws")
.setHandshakeHandler(new DefaultHandshakeHandler(new TomcatRequestUpgradeStrategy()))
.setAllowedOrigins("*")
.withSockJS()
.setSessionCookieNeeded(false)
;
}
}
However, I get a 200 status with a "Welcome to SockJS" message which is generated by TransportHandlingSockJsService in stead of the WebSocketHttpRequestHandler which would generate the upgrade AFAIK
#Configuration
public class WebSocketSecurity extends AbstractSecurityWebSocketMessageBrokerConfigurer{
#Override
protected boolean sameOriginDisabled() {
return true;
}
#Override
protected void configureInbound(MessageSecurityMetadataSourceRegistry messages) {
messages
.nullDestMatcher().permitAll()
.simpSubscribeDestMatchers("/user/queue/errors").permitAll()
.simpDestMatchers("/mobile-server/**").hasRole("ENDUSER")
.simpSubscribeDestMatchers("/user/**", "/topic/**").hasRole("ENDUSER")
.anyMessage().denyAll();
}
}
When I change the config to
#Override
public void configureStompEndpoints(StompEndpointRegistry registry) {
registry
.addEndpoint("/ws")
.setHandshakeHandler(new DefaultHandshakeHandler(new TomcatRequestUpgradeStrategy()))
.setAllowedOrigins("*");
}
to my surprise a call to /ws does lead to a connection upgrade 101. I'm surprised, since the documentation and all examples uniformly use the withSockJS() and the start of any websocket connection AFAIK is a request upgrade.
I can choose to force the upgrade by connecting to /ws/websocket (also not documented). So, I'm not sure what is best.
Any suggestions?
This is expected behavior. It's how the SockJS protocol works:
http://sockjs.github.io/sockjs-protocol/sockjs-protocol-0.3.3.html. There is an initial "greeting" request and then the client starts trying transports one at a time.