Hi Oleg,
thanks for the fast answer!
I adapted the config removing the interceptors, but that doesn't change
the behavior.
See the config below. I also added the source of the two "creates",
which are also used
for the config.
(And what wonders me is, that it works for "http" but not for "https").
best regards
Achim
------------------------------------------------------------------------------------------------
public static CloseableHttpAsyncClient createClient(Configuration
config) {
int connectionIdleSecs =
config.getTimeAsInt(Proxy2Config.HTTP_CONNECTION_IDLE_TIMEOUT,
TimeUnit.SECONDS);
final CloseableHttpAsyncClient client =
HttpAsyncClientBuilder.create()
.disableCookieManagement()
.setDefaultRequestConfig(createCustomRequestConfig(config))
.setConnectionManager(createPoolingConnManager(config))
.setVersionPolicy(HttpVersionPolicy.NEGOTIATE)
.setIOReactorConfig(
IOReactorConfig.custom().setSoTimeout(Timeout.ofSeconds(connectionIdleSecs)).build())
.setKeepAliveStrategy(new
DefaultConnectionKeepAliveStrategy() {
@Override
public TimeValue getKeepAliveDuration(HttpResponse
response, HttpContext context) {
TimeValue keepAlive =
super.getKeepAliveDuration(response, context);
if (keepAlive == null ||
keepAlive.getDuration() < 0) {
// Keep connections alive if a keep-alive value
// has not be explicitly set by the server
keepAlive = KEEP_ALIVE;
}
return keepAlive;
}
}).build();
client.start();
return client;
}
/**
* Create the http request-config.
*
* @param config configuration for the http client
* @return http request-config
* @since 3.0 (changed parameter to Configuration)
*/
private static RequestConfig
createCustomRequestConfig(Configuration config) {
long connecTimeoutMillis =
config.get(Proxy2Config.HTTP_CONNECT_TIMEOUT, TimeUnit.MILLISECONDS);
return
RequestConfig.custom().setConnectionRequestTimeout(Timeout.ofMilliseconds(connecTimeoutMillis
* 4))
.setConnectTimeout(Timeout.ofMilliseconds(connecTimeoutMillis)).build();
}
/**
* Create pooling connection Manager.
*
* @param config configuration for the http client
* @return pooling connection Manager
* @since 3.0 (changed parameter to Configuration)
*/
private static PoolingAsyncClientConnectionManager
createPoolingConnManager(Configuration config) {
long connectionIdleSecs =
config.get(Proxy2Config.HTTP_CONNECTION_IDLE_TIMEOUT,
TimeUnit.MILLISECONDS);
return PoolingAsyncClientConnectionManagerBuilder.create()
.setPoolConcurrencyPolicy(PoolConcurrencyPolicy.STRICT).setConnPoolPolicy(PoolReusePolicy.LIFO)
.setConnectionTimeToLive(TimeValue.ofSeconds(connectionIdleSecs)).setMaxConnTotal(250)
.setMaxConnPerRoute(50).build();
}
------------------------------------------------------------------------------------------------
Am 07.09.22 um 08:57 schrieb Oleg Kalnichevski:
On 9/7/2022 7:49 AM, Achim Kraus wrote:
Hi list,
sometime ago I updated the coap-to-http proxy of Eclipse/Californium
to apache http client 5.1.
That still runs nice using java 7.
But using java 8 or java 11, a "https" request fails with
"ProtocolException: Header 'Connection: keep-alive' is illegal for
HTTP/2 messages "
(see the log below). "http" requests are working.
What wonders me, if the request is HTTP/1.1, why is the response
HTTP/2.0?
If I switch the HttpVersionPolicy.NEGOTIATE to FORCE_HTTP_1 it works
again also for "https".
I updated to 5.1.3, but that doesn't change it.
I use the code in
https://github.com/eclipse-californium/californium/blob/main/californium-proxy2/src/main/java/org/eclipse/californium/proxy2/http/HttpClientFactory.java#L96-L122
to setup the client.
-------------------------------------
HttpAsyncClientBuilder.create().disableCookieManagement()
.setDefaultRequestConfig(createCustomRequestConfig(config))
.setConnectionManager(createPoolingConnManager(config)).setVersionPolicy(HttpVersionPolicy.NEGOTIATE)
.setIOReactorConfig(
IOReactorConfig.custom().setSoTimeout(Timeout.ofSeconds(connectionIdleSecs)).build())
.addRequestInterceptorFirst(new
RequestConnControl()).addRequestInterceptorFirst(new RequestDate())
.addRequestInterceptorFirst(new RequestExpectContinue())
.addRequestInterceptorFirst(new
RequestTargetHost()).addRequestInterceptorFirst(new RequestUserAgent())
.setKeepAliveStrategy(new DefaultConnectionKeepAliveStrategy() {
@Override
public TimeValue getKeepAliveDuration(HttpResponse response,
HttpContext context) {
TimeValue keepAlive =
super.getKeepAliveDuration(response, context);
if (keepAlive == null || keepAlive.getDuration() < 0) {
// Keep connections alive if a keep-alive value
// has not be explicitly set by the server
keepAlive = KEEP_ALIVE;
}
return keepAlive;
}
}).build();
-------------------------------------
Removing the getKeepAliveDuration doesn't help.
I spend also a couple of hours to search the web, but in the flood of
information I was not able to solve this issue.
Hi Achim
You are breaking the H2 protocol processing by adding HTTP/1.1
specific interceptors to the inteceptor chain.
Please remove RequestConnControl, RequestExpectContinue,
RequestTargetHost and RequestUserAgent interceptors from the
initialization code and the problem should go away. Those interceptors
are introduced to the protocol processing chain automatically.
Oleg
---------------------------------------------------------------------
To unsubscribe, e-mail: httpclient-users-unsubscr...@hc.apache.org
For additional commands, e-mail: httpclient-users-h...@hc.apache.org
