So this was working on this project a few months ago. I'm using Google App Engine's Channel API to push messages to my GWT app. I'm using http://code.google.com/p/gwt-gae-channel/ to interact through GWT.
Lets say I send 1 message to the client: "First Message"
The client will receive the message, "First Message" just fine.
Then let's say I send another message, "Second Message" to the client.
The client will again receive the message, "First Message".
This will continue happening. There have been some instances where I'll receive the second message, and it will be the message that gets stuck repeating.
When I finally close the page, and thus close the channel, I again receive the repeated message without sending something from the server.
Does anyone have any idea what is going on? I don't think this was happening when I was working on this a few months ago, and I can see no changes to the GAE Channel API.
Here is some code:
String json = AutoBeanHelper.toJson(proxy);
log.fine("Item's JSON Received: " + json);
List<ChannelEntity> channels = channelDAO.getByUserId();
if (channels.size() > 1) {
log.warning("Multiple channels for single user detected.");
}
ChannelService channelService = ChannelServiceFactory.getChannelService();
for (ChannelEntity channel : channels) {
channelService.sendMessage(new ChannelMessage(channel.getClientId(), json));
}
So whenever I store a new item of a specific type (this is in that entities update function):
1. I turn it into JSON.
2. I then log that JSON.
3. I get that users channel.
4. I send it to that users channel.
When I look at my logs, I see that the variable above that I'm logging is showing correctly, meaning I'm logging the correct JSON message but when I display the JSON in an alert on the client-side as soon as it gets to the client, it's the previous message that seems to be stuck repeating. I really don't see what I could be doing wrong here.
Let me know if you would like to see another part of the code. For good measure, here is the code on the client:
eventBus.addHandler(ReceivedChannelTokenEvent.TYPE, new ReceivedChannelTokenEventHandler() {
#Override
public void onEvent(ReceivedChannelTokenEvent event) {
ChannelFactory.createChannel(event.getChannelToken(), new ChannelCreatedCallback() {
#Override
public void onChannelCreated(Channel channel) {
final Socket channelSocket = channel.open(new SocketListener() {
#Override
public void onOpen() {
Window.alert("Channel Opened");
}
#Override
public void onMessage(String json) {
Window.alert(json);
eventBus.fireEvent(new MessageReceivedEvent(json));
}
#Override
public void onError(SocketError error) {
Window.alert("Channel Error: " + error.getDescription());
if ( error.getDescription().equals(CHANNEL_ERROR_TOKEN_TIME_OUT) ) {
eventBus.fireEvent(new ChannelTimeOutEvent());
}
}
#Override
public void onClose() {
Window.alert("Channel Closed.");
}
});
Window.addWindowClosingHandler(new Window.ClosingHandler() {
#Override
public void onWindowClosing(ClosingEvent event) {
channelSocket.close();
}
});
}
});
}
});
I see a lot of questions on SO where someone has a bug in their code but think it's part of the framework. Without seeing any code, I suspect there's some bug where you think you're sending "Second Message", but you're really sending a cached version of "First Message".
So I was finally able to figure it out. It seems that in the onMessage function within the app when I call
eventBus.fireEvent(new MessageReceivedEvent(json));
it seems that it never returns from this, thus never exiting the onMessage function in the code, causing me to receive the same message repetitively.
Related
I am trying to update a document in MongoDB but cannot get to checking updated status and responding back to user. Below is my code:
#Autowired
ReactiveMongoTemplate mongoTemplate;
public Mono<String> updateUser(UserIn userIn) {
UserResponse resp = new UserResponse();
mongoTemplate.findAndModify(query, update, User.class)
//.doOnSuccess(bsItem -> {
.flatMap(user -> {
if(user.getItemId().equals(userIn.getId("_id")))
resp.setStatus("Updated");
else
resp.setStatus("Failed");
return Mono.just(resp);
}).subscribe();
return Mono.just(resp.getStatus());
}
Even though update is happening in mongodb, it throws NPE while returning. How to get the control after reactor operator is executed here?
You should almost never subscribe in your own application.
The subscriber is the client that initiated the call in this case it is probably the web application. You application is just relaying the data, so your application is a publisher which means you should not subscribe. The web app subscribes.
Try this.
#Autowired
ReactiveMongoTemplate mongoTemplate;
public Mono<String> updateUser(UserIn userIn) {
return mongoTemplate.findAndModify(query, update, User.class)
.flatMap(user -> {
final UserResponse resp = new UserResponse();
if(user.getItemId().equals(userIn.getId("_id")))
resp.setStatus("Updated");
else
resp.setStatus("Failed");
return Mono.just(resp.getStatus());
});
}
A mono is not like a stream, you fetch, map and return, all in the same mono, like a chain of events. An event chain.
Starting off with pubsub. When reading the google cloud documentation, i ran into a snippet of code, and i think i see a flaw with the example.
This is the code i am talking about. It uses the async subscriber.
public class SubscriberExample {
private static final String PROJECT_ID = ServiceOptions.getDefaultProjectId();
private static final BlockingQueue<PubsubMessage> messages = new LinkedBlockingDeque<>();
static class MessageReceiverExample implements MessageReceiver {
#Override
public void receiveMessage(PubsubMessage message, AckReplyConsumer consumer) {
messages.offer(message);
consumer.ack();
}
}
public static void main(String... args) throws Exception {
String subscriptionId = args[0];
ProjectSubscriptionName subscriptionName = ProjectSubscriptionName.of(
PROJECT_ID, subscriptionId);
Subscriber subscriber = null;
try {
subscriber =
Subscriber.newBuilder(subscriptionName, new MessageReceiverExample()).build();
subscriber.startAsync().awaitRunning();
while (true) {
PubsubMessage message = messages.take();
processMessage(message);
}
} finally {
if (subscriber != null) {
subscriber.stopAsync();
}
}
}
My question is, what if a bunch of messages have been acknowledged, and the BlockingQueue is not empty, and the server crashes. Then i would lose some messages right? (Acknowledged in PubSub, but not actually processed).
Wouldn't the best implementation be to only acknowledge the message after the it has been processed? Instead of acknowledging it and leaving it on a queue, and assuming it will be processed. I understand this will decouple the receiving of messages and process of messages, and potentially increase throughput, but still it risks losing messages right?
Yes, one should not acknowledge a message until it has been fully processed. Otherwise, the message may never be processed because it will not be redelivered in the event of a crash or restart if it was acknowledged. I have entered an issue to update the example.
Having trouble sending a large number of bytes from CN1 client to Server. Code below works for anything less than ~7200 bytes. However at byte 7240 the bytes received by the server becomes all zeroes. Not sure why this is occurring, as the connection is still live. The server can still push bytes back just unsure why the client cannot send more data than ~7200. Any advice would be greatly appreciated. Thanks!
if (!outgoingMessages.isEmpty()) {
Socket.connect("my.server.net", 1234, new SocketConnection() {
#Override
public void connectionEstablished(InputStream inStream, OutputStream outStream) {
try {
byte[] msgBytes = outgoingMessages.pop();
outStream.write(ByteConvertor.intToBytes(msgBytes.length));
outStream.write(msgBytes);
} catch(Exception err) {
err.printStackTrace();
}
}
#Override
public void connectionError(int errorCode, String message) {
System.out.println("Error");
}
});
}
Clarification: This is specific to com.codename1.io.Socket and not Steve's CN1Sockets lib
I think this might have been fixed by this pull request, its not available in the simulator builds yet but should work in the source/device. We should update the plugin this week at some point so code should hopefully "just work".
I am getting a strange situation at the code below which simply routes request to Google and returns response.
It works well but when I activate the line commented out as "//Activating this line causes empty response on browser" to print out returned response from http endpoint (Google), response is disappear, nothing is displayed on browser. I thought it might be related with input stream of http response which can be consumed only once and I activated Stream Caching on context but nothing changed.
Apache Camel version is 2.11.0
Any suggestions are greatly appreciated, thanks in advance.
public class GoogleCaller {
public static void main(String[] args) throws Exception {
CamelContext context = new DefaultCamelContext();
context.addRoutes(new RouteBuilder() {
public void configure() {
from("jetty:http://0.0.0.0:8081/myapp/")
.to("jetty://http://www.google.com?bridgeEndpoint=true&throwExceptionOnFailure=false")
.process(new Processor() {
public void process(Exchange exchange) throws Exception {
System.out.println("Response received from Google, is streamCaching = " + exchange.getContext().isStreamCaching());
System.out.println("----------------------------------------------IN MESSAGE--------------------------------------------------------------");
System.out.println(exchange.getIn().getBody(String.class));
System.out.println("----------------------------------------------OUT MESSAGE--------------------------------------------------------------");
//System.out.println(exchange.getOut().getBody(String.class)); //Activating this line causes empty response on browser
}
});
}
});
context.setTracing(true);
context.setStreamCaching(true);
context.start();
}
}
As you use a custom processor to process the message, you should keep it in mind the in message of the exchange has response message from the google, if you are using exchange.getOut(), camel will create a new empty out message for you and treat it as response message.
Because you don't set the out message body in the processor, it makes sense that you get the empty response in the browser.
I have a Windows Phone 7 application built with silverlight. This application has been deployed. I've noticed in the log files that occasionally, my user's actions throw a "NotSupportedException". I have not been able to produce this. However, because of my logs, I know that it is happening in the Execute method shown here:
public void Execute()
{
try
{
// 1. Build the query
string serviceUrl = GetServiceUrl;
// 2. Asynchronously execute the query using HttpWebRequest
WebRequest request = HttpWebRequest.Create(serviceUrl);
request.BeginGetResponse(new AsyncCallback(ServiceRequest_Completed), request);
} catch (Exception ex)
{
LogException(ex, "1");
}
}
private void ServiceRequest_Completed(IAsyncResult result)
{
try
{
// 1. Get the response from the service call
WebRequest request = (WebRequest)(result.AsyncState);
WebResponse response = request.EndGetResponse(result);
// 2. Do stuff with response
}
catch (Exception ex)
{
LogException(ex, "2");
}
}
I know it is happening in the Execute method because the "1" is written in the log file instead of the "2" My question is, what would cause this? I looked at the MSDN documentation and it looks like I'm doing what I should be doing. Like I said, I can't reproduce it locally. But I do know that it is happening regularly by different users because of the log files.
There is a previous question with a very similar title - https://stackoverflow.com/questions/4053197/httpwebrequest-leads-me-to-system-notsupportedexception
The answer to that problem seems to have been using ServiceRequest_Completed instead of new AsyncCallback(ServiceRequest_Completed)