Hi Oleg,
Below my test with just the interesting part of the traces:
1. PoolingAsyncClientConnectionManager with LAX concurrency policy,
max_connections = 1 and pool_timeot = 1. Sending 2 simultaneous requests:
08:34:28.195 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalHttpAsyncExecRuntime
acquireEndpoint 103 ex-0000000002 acquiring endpoint (1
MILLISECONDS)
08:34:28.195
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager lease
264 ex-0000000002 endpoint lease request (1 MILLISECONDS) [route:
{}->http://receiver01:8080][total available: 0; route allocated: 0 of 1; total
allocated: 0 of 0]
08:34:28.197
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 323 ex-0000000002 endpoint leased [route:
{}->http://receiver01:8080][total available: 0; route allocated: 1 of 1; total
allocated: 1 of 1]
08:34:28.197
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 327 ex-0000000002 acquired ep-0000000001
08:34:28.197 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalHttpAsyncExecRuntime$1
completed 117 ex-0000000002 acquired endpoint ep-0000000001
08:34:28.198 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalHttpAsyncExecRuntime
acquireEndpoint 103 ex-0000000001 acquiring endpoint (1
MILLISECONDS)
08:34:28.198
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager lease
264 ex-0000000001 endpoint lease request (1 MILLISECONDS) [route:
{}->http://receiver01:8080][total available: 0; route allocated: 1 of 1; total
allocated: 1 of 1]
( . . .)
08:48:51.451
org.apache.hc.client5.http.impl.async.InternalAbstractHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalAbstractHttpAsyncClient$1 failed
319 ex-0000000003 request failed: Deadline:
2023-06-26T08:48:49.445+0000, -2006 MILLISECONDS overdue
08:48:51.451
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
413 ep-0000000003 connection released [route:
{}->http://receiver01:8080][total available: 1; route allocated: 1 of 1; total
allocated: 1 of 1]
It's OK, 1 request is sent OK and the another one fails because the
max_connections was exceeded.
2. Now I modify the max_connections from 1 to 1000 on runtime. Sending 2
simultaneous requests:
08:36:30.093
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager lease
264 ex-0000000004 endpoint lease request (1 MILLISECONDS) [route:
{}->http://receiver01:8080][total available: 0; route allocated: 0 of 1; total
allocated: 0 of 1]
08:36:30.093
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 323 ex-0000000004 endpoint leased [route:
{}->http://receiver01:8080][total available: 0; route allocated: 1 of 1; total
allocated: 1 of 1]
08:36:30.093
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager lease
264 ex-0000000003 endpoint lease request (1 MILLISECONDS) [route:
{}->http://receiver01:8080][total available: 0; route allocated: 0 of 1; total
allocated: 0 of 1]
( . . . )
08:36:32.099
org.apache.hc.client5.http.impl.async.InternalAbstractHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalAbstractHttpAsyncClient$1 failed
319 ex-0000000003 request failed: Deadline:
2023-06-26T08:36:30.094+0000, -2005 MILLISECONDS overdue
08:36:32.099
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
413 ep-0000000002 connection released [route:
{}->http://receiver01:8080][total available: 1; route allocated: 1 of 1; total
allocated: 1 of 1]
Error, this update from 1 to 1000 is not reflected on runtime.
3. With the same parameters of point 2, now I restart the pool. Sending 2
simultaneous requests:
08:38:46.096
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager lease
264 ex-0000000002 endpoint lease request (1 MILLISECONDS) [route:
{}->http://receiver01:8080][total available: 0; route allocated: 0 of 1000;
total allocated: 0 of 0]
08:38:46.096
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager lease
264 ex-0000000001 endpoint lease request (1 MILLISECONDS) [route:
{}->http://receiver01:8080][total available: 0; route allocated: 0 of 1000;
total allocated: 0 of 0]
08:38:46.096
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 323 ex-0000000002 endpoint leased [route:
{}->http://receiver01:8080][total available: 0; route allocated: 2 of 1000;
total allocated: 2 of 1000]
08:38:46.096
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 327 ex-0000000002 acquired ep-0000000001
08:38:46.096
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 323 ex-0000000001 endpoint leased [route:
{}->http://receiver01:8080][total available: 0; route allocated: 2 of 1000;
total allocated: 2 of 1000]
08:38:46.096
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 327 ex-0000000001 acquired ep-0000000002
( . . . )
08:38:48.123
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
413 ep-0000000001 connection released [route:
{}->http://receiver01:8080][total available: 1; route allocated: 2 of 1000;
total allocated: 2 of 1000]
08:38:48.123 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.HttpAsyncClientProtocolNegotiationStarter$1
onExchangeComplete 150 c-0000000000 Connection is kept
alive
08:38:48.124
org.apache.hc.client5.http.impl.async.InternalAbstractHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalAbstractHttpAsyncClient$1
completed 306 ex-0000000001 message exchange successfully
completed
08:38:48.124 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalHttpAsyncExecRuntime
releaseEndpoint 157 ep-0000000002 releasing valid endpoint
08:38:48.124
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
388 ep-0000000002 releasing endpoint
08:38:48.124
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
404 ep-0000000002 connection c-0000000001 can be kept alive for 100
MILLISECONDS
08:38:48.125
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
413 ep-0000000002 connection released [route:
{}->http://receiver01:8080][total available: 2; route allocated: 2 of 1000;
total allocated: 2 of 1000]
08:38:48.125 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.HttpAsyncClientProtocolNegotiationStarter$1
onExchangeComplete 150 c-0000000001 Connection is kept
alive
It's OK.
Now I was wondering if all this stuff also fails if I change the concurrency
policy from LAX to STRICT. Below the results:
4. Same parameters of point 2 and 3 and pool restarted:
08:47:40.886 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalHttpAsyncExecRuntime
acquireEndpoint 103 ex-0000000001 acquiring endpoint (1
MILLISECONDS)
08:47:40.886 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalHttpAsyncExecRuntime
acquireEndpoint 103 ex-0000000002 acquiring endpoint (1
MILLISECONDS)
08:47:40.887
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager lease
264 ex-0000000001 endpoint lease request (1 MILLISECONDS) [route:
{}->http://receiver01:8080][total available: 0; route allocated: 0 of 1000;
total allocated: 0 of 1000]
08:47:40.887
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager lease
264 ex-0000000002 endpoint lease request (1 MILLISECONDS) [route:
{}->http://receiver01:8080][total available: 0; route allocated: 0 of 1000;
total allocated: 0 of 1000]
08:47:40.888
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 323 ex-0000000001 endpoint leased [route:
{}->http://receiver01:8080][total available: 0; route allocated: 2 of 1000;
total allocated: 2 of 1000]
08:47:40.888
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 323 ex-0000000002 endpoint leased [route:
{}->http://receiver01:8080][total available: 0; route allocated: 2 of 1000;
total allocated: 2 of 1000]
08:47:40.888
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 327 ex-0000000002 acquired ep-0000000001
08:47:40.888
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 327 ex-0000000001 acquired ep-0000000002
( . . . )
08:47:42.911
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
413 ep-0000000002 connection released [route:
{}->http://receiver01:8080][total available: 1; route allocated: 2 of 1000;
total allocated: 2 of 1000]
08:47:42.911 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.HttpAsyncClientProtocolNegotiationStarter$1
onExchangeComplete 150 c-0000000001 Connection is kept
alive
08:47:42.912
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
413 ep-0000000001 connection released [route:
{}->http://receiver01:8080][total available: 2; route allocated: 2 of 1000;
total allocated: 2 of 1000]
08:47:42.912 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.HttpAsyncClientProtocolNegotiationStarter$1
onExchangeComplete 150 c-0000000000 Connection is kept
alive
It's OK.
5. Now I modify the max_connections from 1000 to 1 on runtime. Sending 2
simultaneous requests:
08:48:49.444
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager lease
264 ex-0000000004 endpoint lease request (1 MILLISECONDS) [route:
{}->http://receiver01:8080][total available: 0; route allocated: 0 of 1; total
allocated: 0 of 1]
08:48:49.444
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager lease
264 ex-0000000003 endpoint lease request (1 MILLISECONDS) [route:
{}->http://receiver01:8080][total available: 0; route allocated: 0 of 1; total
allocated: 0 of 1]
08:48:49.444
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
leaseCompleted 323 ex-0000000004 endpoint leased [route:
{}->http://receiver01:8080][total available: 0; route allocated: 1 of 1; total
allocated: 1 of 1]
( . . . )
08:48:51.449 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalHttpAsyncExecRuntime
releaseEndpoint 157 ep-0000000003 releasing valid endpoint
08:48:51.449
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
388 ep-0000000003 releasing endpoint
08:48:51.449
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
404 ep-0000000003 connection c-0000000002 can be kept alive for 100
MILLISECONDS
08:48:51.451
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager$3$1
failed 335 ex-0000000003 endpoint lease failed
08:48:51.451
org.apache.hc.client5.http.impl.async.InternalAbstractHttpAsyncClient
org.apache.hc.client5.http.impl.async.InternalAbstractHttpAsyncClient$1 failed
319 ex-0000000003 request failed: Deadline:
2023-06-26T08:48:49.445+0000, -2006 MILLISECONDS overdue
08:48:51.451
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager
org.apache.hc.client5.http.impl.nio.PoolingAsyncClientConnectionManager release
413 ep-0000000003 connection released [route:
{}->http://receiver01:8080][total available: 1; route allocated: 1 of 1; total
allocated: 1 of 1]
08:48:51.451 org.apache.hc.client5.http.impl.async.InternalHttpAsyncClient
org.apache.hc.client5.http.impl.async.HttpAsyncClientProtocolNegotiationStarter$1
onExchangeComplete 150 c-0000000002 Connection is kept
alive
So it works perfectly, 1 request sent OK and the another fails because
max_connections was exceeded.
It seems that changing the maxConnectionPerRoute does not work on runtime ONLY
when the concurrency policy is LAX.
Anyways maybe I'm doing something wrong ... just let me know what you think.
Thanks,
Joan.
-----Original Message-----
From: Oleg Kalnichevski <[email protected]>
Sent: Friday, June 23, 2023 12:04 PM
To: HttpClient User Discussion <[email protected]>
Subject: Re: Question about max connections
On Fri, 2023-06-23 at 11:52 +0200, Joan grupoventus wrote:
> Hi Oleg,
>
> In the lax mode the pool manager supports the per route max limit, but
> not total max limit. The total max setting has no effect.
> --> Thanks, I can see that when I print the values, the total max is
> always 0.
>
> But the question is if I can modify the per route max limit on the
> fly, or after modifying it I need to restart the pool to apply the new
> value.
>
One can modify the per route max at runtime.
Oleg
> Thanks,
>
> Joan.
>
> -----Original Message-----
> From: Oleg Kalnichevski <[email protected]>
> Sent: Friday, June 23, 2023 11:28 AM
> To: HttpClient User Discussion <[email protected]>
> Subject: Re: Question about max connections
>
> On Fri, 2023-06-23 at 11:23 +0200, Joan grupoventus wrote:
> > Hello,
> >
> > We have an async http pool with LAX concurrency (httpclient 5.2.1):
> > this.phccm =
> > PoolingAsyncClientConnectionManagerBuilder.create().setPoolConcurre
> > nc
> > yPolicy(PoolConcurrencyPolicy.LAX).build();
> >
> > We set the max connections in this way:
> > public void setMaxConnections(int maxConnections) {
> > this.phccm.setMaxTotal(maxConnections);
> > this.phccm.setDefaultMaxPerRoute(maxConnections);
> > }
> >
> > I'm not sure if we are having an issue with this, but before
> > starting to analyse http traces just one question: will it work to
> > call the 'setMaxConnections' method to modify these parameters on
> > the fly, or must I restart the pool to apply the new values?
> >
> > Thanks,
> >
> > Joan.
> >
>
> Hi Joan
>
> In the lax mode the pool manager supports the per route max limit, but
> not total max limit. The total max setting has no effect.
>
> Oleg
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
