[
https://issues.apache.org/jira/browse/HTTPCLIENT-2155?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17331270#comment-17331270
]
Oleg Kalnichevski commented on HTTPCLIENT-2155:
-----------------------------------------------
> Can you please help me understand why this decision was taken that if we are
> not fully consuming the httpResponse the client is closing the connection by
> default instead of releasing it ?
[~saksham2405] There are situations when it is cheaper to close and re-open the
connection rather then trying to read toward the end of the response message.
In some cases servers may even be sending an infinite content stream.
HttpClient leaves the decision whether or not the response content should be
fully consumed up to the user.
Oleg
> CloseableHttpClient leaving too many TIME_WAIT connections
> ----------------------------------------------------------
>
> Key: HTTPCLIENT-2155
> URL: https://issues.apache.org/jira/browse/HTTPCLIENT-2155
> Project: HttpComponents HttpClient
> Issue Type: Bug
> Reporter: Saksham
> Priority: Major
>
> We are using *CloseableHttpClient* with *PoolingHttpClientConnectionManager,*
> following is the configuration:
> {code:java}
> RequestConfig requestConfig = RequestConfig.custom()
> .setConnectTimeout(1000)
> .setConnectionRequestTimeout(1000)
> .setSocketTimeout(1000)
> .build();
> PoolingHttpClientConnectionManager connectionManager = new
> PoolingHttpClientConnectionManager();
> connectionManager.setMaxTotal(100);
> connectionManager.setDefaultMaxPerRoute(100);
> this.httpClient = HttpClients.custom()
> .setDefaultRequestConfig(requestConfig)
> .setConnectionManager(connectionManager)
> .build();
> {code}
>
> Now when we are not consuming the response entity and just relying on
> closeableHttpResponse to release the resources, we end up with 25K
> connections in CLOSE_WAIT state with just over 400RPS to the system. Due this
> system runs out of resources very quickly and requests start failing.
> Problem Snippet:
> {code:java}
> try (CloseableHttpResponse response = this.httpClient.execute(httpGet)){
> if (response.getStatusLine().getStatusCode() == HttpStatus.SC_OK) {
> return EntityUtils.toByteArray(response.getEntity());
> } else {
> if (response.getStatusLine().getStatusCode() >= 500) {
> throw new RemoteCallFailException();
> }
> }
> } catch (IOException e) {
> logger.error("Remote db read request failed [{}]", e.getMessage(), e);
> throw new RemoteCallFailException();
> }
> {code}
> After using *EntityUtils.toByteArray(response.getEntity()) in the else
> block* of above code connections are properly getting released and reused in
> subsequent requests.
> Working Snippet:
> {code:java}
> try (CloseableHttpResponse response = this.httpClient.execute(httpGet)) {
> if (response.getStatusLine().getStatusCode() == HttpStatus.SC_OK) {
> return EntityUtils.toByteArray(response.getEntity());
> } else {
> // Consuming stream to release connection
> EntityUtils.toByteArray(response.getEntity());
> if (response.getStatusLine().getStatusCode() >= 500) {
> throw new RemoteCallFailException();
> }
> }
> } catch (IOException e) {
> logger.error("Remote db read request failed [{}]", e.getMessage(), e);
> throw new RemoteCallFailException();
> }
> {code}
> Can someone please point why releasing of connection is not automatically
> done by closeablehttpResponse.close() ? ( Or maybe point to something wrong
> in the code)
> MVN dependency used for same:
> <dependency>
> <groupId>org.apache.httpcomponents</groupId>
> <artifactId>httpclient</artifactId>
> <version>4.3.3</version>
> </dependency>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
