I am studying Spring webflux and faced with a situation that I am not understanding.
Spring version = 2.2.4.RELEASE
My computer has 4 cores.
My Class:
#RestController
#RequestMapping("rest/v1/")
public class RootController {
private final WebClient webClient =
WebClient.create("http://localhost:8081");
#GetMapping("/booking")
public Mono book() {
return Mono.just("A");
}
#GetMapping("/sleep")
public Mono<String> sleep(#RequestParam("time") long time) throws InterruptedException {
System.out.printf("Thread name is %s, date is %s. \n", Thread.currentThread().getName(), new Date());
String uri = "/rest/v1/sleep?time=" + time;
String uuid = UUID.randomUUID().toString();
System.out.println(">>>>>>>>>>>>" + uuid);
Mono<String> stringMono = webClient.get().uri(uri)
.retrieve()
.bodyToMono(String.class);
System.out.printf("Finish -> Thread name is %s, date is %s. \n", Thread.currentThread().getName(), new Date());
return stringMono;
}
}
I have another application in my machine, running on port 8081.
This endoint only sleep for a period of time that received by param.
The goal is test how is the webflux behaviour when call a slow endpoint.
I am perfoming a request like this:
http://localhost:8082/rest/v1/sleep?time=20000
In this case, the response will take 20s.
When I perform another request during this period, the controller does not process the request, it keep waiting for this request finish (20s).
Should Spring process another request while is waiting for a network call?
This is the log:
Thread name is reactor-http-epoll-3, date is Sat May 16 19:06:54 BRT 2020.
>>>>>>>>>>>>22b36a46-e36b-4104-981a-e132483a334e
Finish -> Thread name is reactor-http-epoll-3, date is Sat May 16 19:06:54 BRT 2020.
Thread name is reactor-http-epoll-3, date is Sat May 16 19:07:14 BRT 2020.
>>>>>>>>>>>>076a0eec-3bb1-485a-ab88-9b11c6b9a5bd
Finish -> Thread name is reactor-http-epoll-3, date is Sat May 16 19:07:14 BRT 2020.
The problem was a chrome issue that has queued my request like "Martin Tarjányi" said in his comment:
Did you hit your endpoint from Chrome browser? I experienced similar
issue, not sure about the reason, though. As I recall it worked well
either in Firefox or Edge, though. You can also call it
programmatically from another app. Edit:
stackoverflow.com/a/53236741/6051176
Thanks a lot #Martin Tarjányi
Request using other clients or change parameter works well. The same resource chrome will queue request.
Related
We are using AggregatingReplyingKafkaTemplate for sending and recieving a reply of messages.
I want to increase the timeout. But it seems the Kafka Listener is timing out in this case in around 30-40 seconds.
I have tried increasing the default timeout to 10 mins (just for testing ). But it did not help
aggregatingReplyingKafkaTemplate.setDefaultReplyTimeout(Duration.ofSeconds(600)); //changing the default timeout
Code when we are sending the message
private final AggregatingReplyingKafkaTemplate<String, T, R> aggregatingReplyingKafkaTemplate;
public RequestReplyFuture<String, T, Collection<ConsumerRecord<String, R>>> sendAndReceive(String producerTopic, T record, Headers headers) {
ProducerRecord<String, T> producerRecord = new ProducerRecord(producerTopic, record);
addHeaders(producerRecord, headers);
return this.aggregatingReplyingKafkaTemplate.sendAndReceive(producerRecord);
}
Best Regards,
Saurav
I was using DeferredResult for async request with a timeout which was causing my original HTTP request to time out. This HTTP request was internally using ReplyingTemplate to communicate with the downstream services.
I did increase the ReplyingTemplate's default timeout as i mentioned above.
So, its working now.
I want to be able to extract the List<Payload> from the Mono<List<Payload>> to pass it to a downstream service for processing (or maybe return from the read(RequestParams params) method, instead of it returning void):
#PostMapping("/subset")
public void read(#RequestBody RequestParams params){
Mono<List<Payload>> result = reader.read(params.getDate(), params.getAssetClasses(), params.getFirmAccounts(), params.getUserId(), params.getPassword());
....
}
where reader.read(...) is a method on an autowired Spring service utilizing a webClient to get the data from external web service API:
public Mono<List<Payload>> read(String date, String assetClasses, String firmAccounts, String id, String password) {
Flux<Payload> nodes = client
.get()
.uri(uriBuilder -> uriBuilder
.path("/api/subset")
.queryParam("payloads", true)
.queryParam("date", date)
.queryParam("assetClasses", assetClasses)
.queryParam("firmAccounts", firmAccounts)
.build())
.headers(header -> header.setBasicAuth("abc123", "XXXXXXX"))
.retrieve()
.onStatus(HttpStatus::is4xxClientError, response -> {
System.out.println("4xx error");
return Mono.error(new RuntimeException("4xx"));
})
.onStatus(HttpStatus::is5xxServerError, response -> {
System.out.println("5xx error");
return Mono.error(new RuntimeException("5xx"));
})
.bodyToFlux(Payload.class);
Mono<List<Payload>> records = nodes
.collectList();
return records;
}
Doing a blocking result.block() is not allowed in WebFlux and throws an exception:
new IllegalStateException("block()/blockFirst()/blockLast() are blocking, which is not supported in thread ..." ;
What is a proper way to extract the contents of a Mono in WebFlux?
Is it some kind of a subscribe()? What would be the syntax?
Thank you in advance.
There is no "proper way" and that is the entire point. To get the value you need to block, and blocking is bad in webflux for many reasons (that I won't go into right now).
What you should do is to return the publisher all the way out to the calling client.
One of the things that many usually have a hard time understanding is that webflux works with a producer (Mono or Flux) and a subscriber.
Your entire service is also a producer, and the calling client can be seen as the subscriber.
Think of it as a long chain, that starts at the datasource, and ends up in the client showing the data.
A simple rule of thumb is that whomever is the final consumer of the data is the subscriber, everyone else is a producer.
So in your case, you just return the Mono<List<T> out to the calling client.
#PostMapping("/subset")
public Mono<List<Payload>> read(#RequestBody RequestParams params){
Mono<List<Payload>> result = reader.read(params.getDate(), params.getAssetClasses(), params.getFirmAccounts(), params.getUserId(), params.getPassword());
return result;
}
While the following does return the value of the Mono observable in the logs:
#PostMapping("/subset")
#ResponseBody
public Mono<ResponseEntity<List<Payload>>> read1(#RequestBody RequestParams params){
Mono<List<Payload>> result = reader.read(params.getDate(), params.getAssetClasses(), params.getFirmAccounts(), params.getUserId(), params.getPassword());
return result
.map(e -> new ResponseEntity<List<PayloadByStandardBasis>>(e, HttpStatus.OK));
}
the understanding I was seeking was a proper way to compose a chain of calls, with WebFlux, whereby a response from one of its operators/legs (materialized as as a result from a webclient call, producing a set of records, as above) could be passed downstream to another operator/leg to facilitate a side effect of saving those records in a DB, or something to that effect.
It would probably be a good idea to model each of those steps as a separate REST endpoint, and then have another endpoint for a composition operation which internally calls each independent endpoint in the right order, or would other design choices be more preferred?
That is ultimately the understanding I was looking for, so if anyone wants to share an example code as well as opinions to better implement the set of steps described above, I'm willing to accept the most comprehensive answer.
Thank you.
I am new to Reactive programming paradigm, but recently I have decided to base a simple Http client on Spring WebClient, since the old sync RestTemplate is already under maintenance and might be deprecated in upoming releases.
So first I had a look at Spring documentation and, after that, I've searched the web for examples.
I must say that (only for the time being) I have consciously decided not to go through the Reactor lib documentation, so beyond the Publisher-Subscriber pattern, my knowledge about Mono's and Flux's is scarce. I focused instead on having something working.
My scenario is a simple POST to send a callback to a Server from which the client is only interested in response status code. No body is returned. So I finally came up with this code snippet that works:
private void notifyJobSuccess(final InternalJobData jobData) {
SuccessResult result = new SuccessResult();
result.setJobId(jobData.getJobId());
result.setStatus(Status.SUCCESS);
result.setInstanceId(jobData.getInstanceId());
log.info("Result to send back:" + System.lineSeparator() + "{}", result.toString());
this.webClient.post()
.uri(jobData.getCallbackUrl())
.body(Mono.just(result), ReplaySuccessResult.class)
.retrieve()
.onStatus(s -> s.equals(HttpStatus.OK), resp -> {
log.info("Expected CCDM response received with HttpStatus = {}", HttpStatus.OK);
return Mono.empty();
})
.onStatus(HttpStatus::is4xxClientError, resp -> {
log.error("CCDM response received with unexpected Client Error HttpStatus {}. "
+ "The POST request sent by EDA2 stub did not match CCDM OpenApi spec", resp.statusCode());
return Mono.empty();
})
.onStatus(HttpStatus::is5xxServerError, resp -> {
log.error("CCDM response received with unexpected Server Error HttpStatus {}", resp.statusCode());
return Mono.empty();
}).bodyToMono(Void.class).subscribe(Eda2StubHttpClient::handleResponseFromCcdm);
}
My poor understanding of how the reactive WebClient works starts with the call to subscribe. None of the tens of examples that I checked before coding my client included such a call, but the fact is that before I included that call, the Server was sitting forever waiting for the request.
Then I bumped into the mantra "Nothing happens until you subscribe". Knowing the pattern Plublisher-Subscriber I knew that, but I (wrongly) assumed that the subscription was handled by WebClient API, in any of the exchage, or bodyToMono methods... block() definitely must subscribe, because when you block it, the request gets out at once.
So my first question is: is this call to subscribe() really needed?
Second question is why the method StubHttpClient::handleResponse is never called back. For this, the only explanation that I find is that as the Mono returned is a Mono<Void>, because there is nothing in the response besides the status code, as it is never instantiated, the method is totally dummy... I could even replace it by just .subscribe(). Is this a correct assumption.
Last, is it too much to ask for a complete example of a a method receiving a body in a Mono that is later consumed? All examples I find just focus on getting the request out, but how the Mono or Flux is later consumed is now beyond my understanding... I know that I have to end up checking the Reactor doc sooner better than later, but I would appreciate a bit of help because I am having issues with Exceptions and errors handlin.
Thanks!
Some time has passed since I asked for help here. Now I'd like not to edit but to add an answer to my previous question, so that the answer remains clear and separate from he original question and comments.
So here goes a complete example.
CONTEXT: An application, acting as a client, that requests an Access Token from an OAuth2 Authorization server. The Access Token is requested asynchronously to avoid blocking the appliction's thread while the token request is processed at the other end and the response arrives.
First, this is a class that serves Access Token to its clients (method getAccessToken): if the Access Token is already initialized and it's valid, it returns the value stored; otherwise fetches a new one calling the internal method fetchAccessTokenAsync:
public class Oauth2ClientBroker {
private static final String OAUHT2_SRVR_TOKEN_PATH= "/auth/realms/oam/protocol/openid-connect/token";
private static final String GRANT_TYPE = "client_credentials";
#Qualifier("oAuth2Client")
private final WebClient oAuth2Client;
private final ConfigurationHolder CfgHolder;
#GuardedBy("this")
private String token = null;
#GuardedBy("this")
private Instant tokenExpireTime;
#GuardedBy("this")
private String tokenUrlEndPoint;
public void getAccessToken(final CompletableFuture<String> completableFuture) {
if (!isTokenInitialized() || isTokenExpired()) {
log.trace("Access Token not initialized or has exired: go fetch a new one...");
synchronized (this) {
this.token = null;
}
fetchAccessTokenAsync(completableFuture);
} else {
log.trace("Reusing Access Token (not expired)");
final String token;
synchronized (this) {
token = this.token;
}
completableFuture.complete(token);
}
}
...
}
Next, we will see that fetchAccessTokenAsync does:
private void fetchAccessTokenAsync(final CompletableFuture<String> tokenReceivedInFuture) {
Mono<String> accessTokenResponse = postAccessTokenRequest();
accessTokenResponse.subscribe(tr -> processResponseBodyInFuture(tr, tokenReceivedInFuture));
}
Two things happen here:
The method postAccessTokenRequest() builds a POST request and declares how the reponse will be consumed (when WebFlux makes it available once it is received), by using exchangeToMono:
private Mono postAccessTokenRequest() {
log.trace("Request Access Token for OAuth2 client {}", cfgHolder.getClientId());
final URI uri = URI.create(cfgHolder.getsecServiceHostAndPort().concat(OAUHT2_SRVR_TOKEN_PATH));
} else {
uri = URI.create(tokenUrlEndPoint);
}
}
log.debug("Access Token endpoint OAuth2 Authorization server: {}", uri.toString());
return oAuth2Client.post().uri(uri)
.body(BodyInserters.fromFormData("client_id", cfgHolder.getEdaClientId())
.with("client_secret", cfgHolder.getClientSecret())
.with("scope", cfgHolder.getClientScopes()).with("grant_type", GRANT_TYPE))
.exchangeToMono(resp -> {
if (resp.statusCode().equals(HttpStatus.OK)) {
log.info("Access Token successfully obtained");
return resp.bodyToMono(String.class);
} else if (resp.statusCode().equals(HttpStatus.BAD_REQUEST)) {
log.error("Bad request sent to Authorization Server!");
return resp.bodyToMono(String.class);
} else if (resp.statusCode().equals(HttpStatus.UNAUTHORIZED)) {
log.error("OAuth2 Credentials exchange with Authorization Server failed!");
return resp.bodyToMono(String.class);
} else if (resp.statusCode().is5xxServerError()) {
log.error("Authorization Server could not generate a token due to a server error");
return resp.bodyToMono(String.class);
} else {
log.error("Authorization Server returned an unexpected status code: {}",
resp.statusCode().toString());
return Mono.error(new Exception(
String.format("Authorization Server returned an unexpected status code: %s",
resp.statusCode().toString())));
}
}).onErrorResume(e -> {
log.error(
"Access Token could not be obtained. Process ends here");
return Mono.empty();
});
}
The exchangeToMono method does most of the magic here: tells WebFlux to return a Mono that will asynchronously receive a signal as soon as the response is received, wrapped in a ClientResponse, the parameter resp consumed in the lambda. But it is important to keep in mind that NO request has been sent out yet at this point; we are just passing in the Function that will take the ClientResponse when it arrives and will return a Mono<String> with the part of the body of our interest (the Access Token, as we will see).
Once the POST is built and the Mono returned, then the real thing starts when we subscribe to the Mono<String> returned before. As the Reacive mantra says: nothing happens until you subscribe or, in our case, the request is not actually sent until something attempts to read or wait for the response. There are other ways in WebClient fluent API to implicitly subscribe, but we have chosen here the explicit way of returing the Mono -which implements the reactor Publisher interface- and subscribe to it. Here we blocking the thread no more, releasing CPU for other stuff, probably more useful than just waiting for an answer.
So far, so good: we have sent out the request, released CPU, but where the processing will continue whenever the response comes? The subscribe() method takes as an argument a Consumer parameterized in our case with a String, being nothing less than the body of the response we are waiting for, wrapped in Mono. When the response comes, WebFlux will notify the event to our Mono, which will call the method processResponseBodyInFuture, where we finally receive the response body:
private void processResponseBodyInFuture(final String body, final CompletableFuture<String> tokenReceivedInFuture) {
DocumentContext jsonContext = JsonPath.parse(body);
try {
log.info("Access Token response received: {}", body);
final String aTkn = jsonContext.read("$.access_token");
log.trace("Access Token parsed: {}", aTkn);
final int expiresIn = jsonContext.read("$.expires_in");
synchronized (this) {
this.token = aTkn;
this.tokenExpireTime = Instant.now().plusSeconds(expiresIn);
}
log.trace("Signal Access Token request completion. Processing will continue calling client...");
tokenReceivedInFuture.complete(aTkn);
} catch (PathNotFoundException e) {
try {
log.error(e.getMessage());
log.info(String.format(
"Could not extract Access Token. The response returned corresponds to the error %s: %s",
jsonContext.read("$.error"), jsonContext.read("$.error_description")));
} catch (PathNotFoundException e2) {
log.error(e2.getMessage().concat(" - Unexpected json content received from OAuth2 Server"));
}
}
}
The invocation of this method happens as soon as the Mono is signalled about the reception of the response. So here we try to parse the json content with an Access Token and do something with it... In this case call complete() onto the CompletableFuture passed in by the caller of the initial method getAccessToken, that hopefully will know what to do with it. Our job is done here... Asynchronously!
Summary:
To summarize, these are the basic considerations to have your request sent out and the responses processed when you ise reactive WebClient:
Consider having a method in charge of preparing the request by means of the WebClient fluent API (to set http method, uri, headers and body). Remember: by doing this you are not sending any request yet.
Think on the strategy you will use to obtain the Publisher that will be receive the http client events (response or errors). retreive() is the most straight forward, but it has less power to manipulate the response than exchangeToMono.
Subscribe... or nothing will happen.
Many examples you will find around will cheat you: they claim to use WebClient for asyncrhony, but then they "forget" about subscribing to the Publisher and call block() instead. Well, while this makes things easier and they seem to work (you will see responses received and passed to your application), the thing is that this is not asynchronous anymore: your Mono (or Flux, whatever you use) will be blocking until the response arrives. No good.
Have a separate method (being the Consumer passed in the subscribe() method) where the response body is processed.
I am using the spring WebClient to make two calls in parallel.
One of the call results is passed back as a ResponseEntity, and the other result is inspected and then disregarded. Although the transactions are both successful, I see an IllegalReferenceCountException that occurs before any of the WebClient calls actually get executed.
What I see in my logging is that the container logs the exception, then my two HTTP requests get executed successfully, and one of these responses gets returned to the client.
If the shouldBackfill() function returns false, then I execute one HTTP request and return that response (and the IllegalReferenceCountException does not occur).
I was initially thinking that I should release the reference in the second response that I disregard.
If I attempt to call releaseBody() directly on the WebClient response. (See https://docs.spring.io/spring-framework/docs/current/javadoc-api/org/springframework/web/reactive/function/client/ClientResponse.html), this does not help. I assume now that the container is detecting that the WebClient request that I disregarded is in an illegal state, hence the error detection. But what I don't understand is that the actual request occurs AFTER the IllegalReferenceCountException gets logged.
Any ideas here on how to get around this? I am wondering if the exception is actually NOT any kind of leak.
The code looks like this:
fun execute(routeHttpRequest: RouteHttpRequest): Mono<ResponseEntity<String>> =
propertyRepository.getProperty(routeHttpRequest.propertyId.orDefault())
.flatMap {
val status = it.getOrElse { unknownStatus(routeHttpRequest.propertyId.orDefault()) }
val response1 = execute(routeHttpRequest, routingRepository.webClientFor(routeHttpRequest))
if (shouldBackfill(routeHttpRequest, status.type())) {
val response2 =
execute(routeHttpRequest, routingRepository.shadowOrBackfillWebClientFor(routeHttpRequest))
zip(response1, response2).map { response ->
compare(routeHttpRequest, response.t1, response.t2, status.type())
response.t1 // response.t2 is NOT returned here..
}
} else response1
}
// This function returns a wrapper on a spring Webclient that makes an HTTP post.
//
private fun execute(routeHttpRequest: RouteHttpRequest, client: Mono<MyWebClient>) =
client
.flatMap { dataService.execute(routeHttpRequest, it) }
.subscribeOn(Schedulers.elastic()) // TODO: consider a dedicated executor here?
private fun shouldBackfill(routeHttpRequest: RouteHttpRequest, migrationStatus: MigrationStatusType): Boolean {
... this logic returns true when we should execute 2 requests in parallel
}
Here's the exception and partial trace:
io.netty.util.IllegalReferenceCountException: refCnt: 0, decrement: 1
at io.netty.util.internal.ReferenceCountUpdater.toLiveRealRefCnt(ReferenceCountUpdater.java:74)
at io.netty.util.internal.ReferenceCountUpdater.release(ReferenceCountUpdater.java:138)
at io.netty.buffer.AbstractReferenceCountedByteBuf.release(AbstractReferenceCountedByteBuf.java:100)
at io.netty.util.ReferenceCountUtil.release(ReferenceCountUtil.java:88)
Sorry for not posting the exact code. Fix- I was passing the incoming http request org.springframework.core.io.buffer.DataBuffer directly to the WebClient request body. This was intentional because my application is acting as a proxy service. The problem came up when I attempted to make two outbound WebClient calls in parallel - the container was trying to release the underlying buffer twice, and the IllegalReferenceCountException occurs. My fix was to just copy the DataBuffer byte array into a new buffer before sending the request along to it's destination.
I have an rest api with Play 2.3.x on a Docker container. This api speak with PostgresSQL 9.4 database (on a Docker container too) and serves JSON. Client is a isomorphic app with React (on Docker container too).
When client change route, several request is send to the api. The problem is that every request is abnormally slow. For example, I have this uri:
GET /api/taxon/185214/geojson
When I send this request directly since the browser, response is shown at the end of 4s. When this request is send with other request, it takes 12s longer.
What can cause this time ?
Of course, every action of api are async. For example, geojson action:
public static Promise<Result> showGeoJson(final Long id)
{
Promise<GeoJsonModel> promise = F.Promise.promise(
new Function0<GeoJsonModel>() {
public GeoJsonModel apply() {
// Get data with ebean
return geojson;
}
}
);
return promise.map(
new Function<GeoJsonModel, Result>() {
public Result apply(GeoJsonModel geojson) {
// return Result
}
}
);
}
Does each request should be treated in another thread ?
EDIT
So, after several tests, it's just my development computer whish is slow. Indeed, on other computer the response time is normal.