Hi,
I have a list with size of 500k and I have to make a request to a server with hash parameters.
The server accepts JSON Array of 200 objects. so I could send 200 items each time.
But still I need to split the list each time and send that part to server.
I have a method for this which makes the http post request. and I want to use spring boot options (if available) to call the method with different threads and get the response back and merge them into one.
I did it using java CompletableFuture class without any springboot tags. but you Could use #async for your method too. sample code :
var futures = new ArrayList<CompletableFuture<List<Composite>>>();
var executor = Executors.newFixedThreadPool(Runtime.getRuntime().availableProcessors());
for (List<CompositeRecord> records : Lists.partition(recordsList, 200)) {
var future = CompletableFuture.supplyAsync(() -> /* call your method here */, executor);
futures.add(future);
Thread.sleep(2000);
}
CompletableFuture.allOf(futures.toArray(new CompletableFuture[0])).exceptionally(ex -> null).join(); // avoid throwing an exception in the join() call
var futureMap = futures.stream().collect(Collectors.partitioningBy(CompletableFuture::isCompletedExceptionally));
var compositeWithErrorList = new ArrayList<Composite>();
futureMap.get(false).forEach(l -> {
try {
compositeWithErrorList.addAll(l.get());
} catch (InterruptedException | ExecutionException e) {
e.printStackTrace();
}
});
after the code is executed you will have a map of done and undone futures.
Related
I am from the old world that think webapi should return a strong typed object and let json serialization return data.
However, recently we got this requirement:
We have a sql table which has more than 500 columns.
The customer always want to return all the columns.
Our c# code does nothing other than reading the SqlDatareader, convert the reader to a c# object and return result.
In this case, wouldn't better to do this (example copied from another stackoverflow post). Basically just return a stream? Does returning a stream still considered to be anti-pattern?
public HttpResponseMessage SomeMethod(List<string> someIds)
{
HttpResponseMessage resp = new HttpResponseMessage();
resp.Content = new PushStreamContent(async (responseStream, content, context) =>
{
await CopyBinaryValueToResponseStream(responseStream, someIds);
});
return resp;
}
private static async Task CopyBinaryValueToResponseStream(Stream responseStream, int imageId)
{
// PushStreamContent requires the responseStream to be closed
// for signaling it that you have finished writing the response.
using (responseStream)
{
using (SqlConnection connection = new SqlConnection(connectionString))
{
await connection.OpenAsync();
using (SqlCommand command = new SqlCommand("SELECT 500 columns FROM [StupidWideTable] WHERE ....", connection))
{
.....
using (SqlDataReader reader = await command.ExecuteReaderAsync(CommandBehavior.SequentialAccess))
{
if (await reader.ReadAsync())
{
if (!(await reader.IsDBNullAsync(0)))
{
using (Stream data = reader.GetStream(0))
{
// Asynchronously copy the stream from the server to the response stream
await data.CopyToAsync(responseStream);
}
}
}
}
}
}
}// close response stream
}
Does returning a stream still considered to be anti-pattern?
Well, that depends on what you want to do. For example, if you want to return a 500 if the SQL server fails partway through, then you shouldn't return a stream.
Streaming results works fine on ASP.NET, but it's important to note that all headers (including the response status code) are sent before the stream begins. So you'll send an immediate 200 when you start streaming the result, and if there's an error later on there's no way to go back in time and change that to a 500. Or add some kind of Continue header.
In other words, yes it's supported; but you lose all the benefits of model binding, content negotiation, exception handlers, etc., because you're bypassing that whole pipeline.
We have a Controller running on tomcat 8.5.32 which receives a POST request with query params
/{path_param}/issue?title=4&description=5
request body is empty
Then controller redirects this request to Spring Boot microservice with tomcat 9.0.27.
At line
CloseableHttpResponse result = httpClient.execute(request);
request.getURI().getQuery() equals&title=1&description=2
But when it arrives to microservice parameters are duplicated (title=[4,4]&description=[5,5]).
This is the code which redirects request to microservice
private static <T, U> T executePostRequest(String url, U body, HttpServletRequest httpServletRequest, Function<String, T> readValueFunction) {
try (CloseableHttpClient httpClient = HttpClientBuilder.create().build()) {
URIBuilder uriBuilder = new URIBuilder(url);
httpServletRequest.getParameterMap().forEach((k, v) -> Arrays.stream(v).forEach(e -> uriBuilder.addParameter(k, e)));
HttpPost request = new HttpPost(uriBuilder.build());
CloseableHttpResponse result = httpClient.execute(request);
String json = EntityUtils.toString(result.getEntity(), "UTF-8");
handleResultStatus(result, json);
return readValueFunction.apply(json);
} catch (IOException | URISyntaxException e) {
...
}
}
I found that there was similar issue with jetty and it was fixed but did not find anything related to tomcat - and how it can be fixed.
I saw also this topic whith suggestion how to handle duplicated parameters in spring boot but i am wondering if anyone else experienced same issue and how did you resolve it if yes.
It's not a bug, it's a feature present in every servlet container.
The Servlet API does not require for the request parameters to have unique names. If you send a POST request for http://example.com/app/issue?title=1&description=2 with a body of:
title=3&description=4
then each parameter will have multiple values: title will have values 1 and 3, while description will have values 2 and 4 in that order:
Data from the query string and the post body are aggregated into the request
parameter set. Query string data is presented before post body data. For example, if
a request is made with a query string of a=hello and a post body of a=goodbye&a=
world, the resulting parameter set would be ordered a=(hello, goodbye, world).
(Servlet specification, section 3.1)
If you want to copy just the first value of the parameters use:
httpServletRequest.getParameterMap()//
.forEach((k, v) -> uriBuilder.addParameter(k, v[0]));
Staring with the tutorial code at benwilcock/spring-rsocket-demo I am trying to write a server that replicates messages to a second server before responding to a client.
To try to debug my issues I am only attempting a trivial ping-pong exchange between servers. Only when the second server responds to the pong message should the first server reply to the client:
#MessageMapping("request-response")
Mono<Message> requestResponse(final Message request) {
// register a mono that will be completed when replication to another server has happened
String uuid = UUID.randomUUID().toString();
Mono<Message> deferred = Mono.create(sink -> replicationNexus.registerRequest(uuid, sink));
// FIXME attempt to send a nested request-response message that will complete the outer message later
this.requesterMono.flatMap(requester -> requester.route("pong")
.data(uuid)
.retrieveMono(String.class))
.subscribeOn(Schedulers.elastic())
.subscribe( uuid2 -> replicationNexus.complete(uuid2, new Message(SERVER, RESPONSE)));
// return the deferred work that will be completed by the pong response
return deferred;
}
That logic is trying to use this answer to create a connection to the second server that will reconnect:
this.requesterMono = builder.rsocketConnector(connector -> connector
.reconnect(Retry.fixedDelay(Integer.MAX_VALUE, Duration.ofSeconds(1))))
.connectTcp("localhost", otherPort).cache();
To complete the picture here is the trivial ping-pong logic:
#MessageMapping("pong")
public Mono<String> pong(String m) {
return Mono.just(m);
}
and here is the logic that holds the state of the outer client response that is completed when the other server responds:
public class ReplicationNexus<T> {
final Map<String, MonoSink<T>> requests = new ConcurrentHashMap<>();
public void registerRequest(String v, MonoSink<T> sink) {
requests.put(v, sink);
}
public boolean complete(String uuid, T message) {
Optional<MonoSink<T>> sink = Optional.of(requests.get(uuid));
if( sink.isPresent() ){
sink.get().success(message);
}
return sink.isPresent();
}
}
Debugging the second server it never runs the pong method. It seems that the first server does not actually send the inner request message.
What is the correct way to run an inner request-response exchange that completes an outer message exchange with automated reconnection logic?
Not sure if I'm missing some of the complexity of your question, but if the middle server is just activing like a proxy I'd start with the simplest case of chaining through the calls. I feel like I'm missing some nuance of the question, so let's work through that next.
#MessageMapping("runCommand")
suspend fun runCommandX(
request: CommandRequest,
): Mono<String> {
val uuid = UUID.randomUUID().toString()
return requesterMono
.flatMap { requester: RSocketRequester ->
requester.route("pong")
.data("TEST")
.retrieveMono(String::class.java)
}
.doOnSubscribe {
// register request with uuid
}
.doOnSuccess {
// register completion
}
.doOnError {
// register failure
}
}
Generally if you can avoid calling subscribe yourself in typical spring/reactive/rsocket code. You want the framework to do this for you.
I am using web client in a spring application
I am facing memory leak issues while doing the same
I am using below code to get the response body for non 2XX response from service:
return client.get()
.uri(uriString)
.headers(ServiceCommonUtil.getHttpHeaderConsumer(headersMap))
.exchange()
.flatMap(clientResponse -> {
try {
clientResponse.body((clientHttpResponse, context) ->
clientHttpResponse.getBody());
logResponseStatus(clientResponse.statusCode(), serviceName);
return clientResponse.bodyToMono(String.class);
} catch (Exception e) {
return null;
}
})
and later on subscriber uses subscribe/ error block to process this response.
responseMono.subscribe(response -> {
//process response string
},error->{
//process error response
});
My question is, if i use dispose method on responseMono, it takes way long time for processing while without it i face memory leak issues.
Am i doing anything wrong here?
Yes, actually you are not consumming response in case of Exception is thrown.
If you want to use exchange() your responsibillity is to consume response.
See: docs
Take a look on toBodilessEntity()/ releaseBody() in 'ClientResponse` api.
Seems you've gotten a little complicated. Why a try/catch block in the clientResponse lambda? If your logResponseStatus throws a checked exception then handle it there. I suggest starting simpler.
Ex 1:
Mono<String> stringMono = webClient.get().uri("test").header("head", "value").exchange().flatMap(clientResponse->clientResponse.bodyToMono(String.class));
stringMono.subscribe(System.out::println);
Ex 2:
Mono<String> stringMono = webClient.get().uri("test").header("head", "value").exchange().flatMap(clientResponse->clientResponse.body(BodyExtractors.toMono(String.class)));
stringMono.subscribe(System.out::println);
Ex 3:
Mono<String> stringMono = webClient.get().uri("test").header("head", "value").retrieve().bodyToMono(String.class);
stringMono.subscribe(System.out::println);
For logging it is better to use ExchangeFilterFunctions. See How to intercept a request when using SpringBoot WebClient
.
I configured LitteProxy as a reverse proxy and trying to implement a filter. When I send a HTTP GET , the clientToProxyRequest filter adapter is called twice. Once with DefaultHttpRequest as httpObject and the second time with LastHttpContent. It can be something to do with chunks ? How do I do this right ?
HttpProxyServerBootstrap bs = DefaultHttpProxyServer.bootstrap();
//reverse proxy
bs.withAllowRequestToOriginServer(true);
Filter is created as :
new HttpFiltersAdapter(originalRequest) {
#Override
public HttpResponse clientToProxyRequest(HttpObject httpObject) {
//doing filtering here..
return null;
}
}