In addition to my inline notes, another thought came to my mind:

Is it possible to - instead of closing the connection apruptly - maybe set the 
expiry to “now” on those and then have the pool cleaning up? 
The problem is I’m building a client library and I dont want to have a monitor 
thread running (and there is one for the reactor and N workers already running).

Maybe that approach is less destructive? I dont mind if it takes 30 secs or so 
to clean them up, but it needs to be done at some point.


On 16 Dec 2013, at 15:58, Oleg Kalnichevski <[email protected]> wrote:

> On Mon, 2013-12-16 at 15:21 +0100, Michael Nitschinger wrote:
>> A Quick follow up on that one. I now did it like that and it seems to work:
>> 
>>    public void cleanup(final HttpHost host) {
>>      enumAvailable(new PoolEntryCallback<HttpHost, NHttpClientConnection>() {
>>        @Override
>>        public void process(PoolEntry<HttpHost, NHttpClientConnection> entry) 
>> {
>>          if (entry.getRoute().equals(host)) {
>>            entry.close();
>>          }
>>        }
>>      });
>> 
>>      enumLeased(new PoolEntryCallback<HttpHost, NHttpClientConnection>() {
>>        @Override
>>        public void process(PoolEntry<HttpHost, NHttpClientConnection> entry) 
>> {
>>          if (entry.getRoute().equals(host)) {
>>            entry.close();
>>          }
>>        }
>>      });
>>    }
>> 
>> but of course this can pretty much kill the selectors in flight.. that’s 
>> whats coming up sometimes:
>> 
>> java.nio.channels.CancelledKeyException
>>      at sun.nio.ch.SelectionKeyImpl.ensureValid(SelectionKeyImpl.java:73)
>>      at sun.nio.ch.SelectionKeyImpl.interestOps(SelectionKeyImpl.java:77)
>>      at 
>> org.apache.http.impl.nio.reactor.IOSessionImpl.getEventMask(IOSessionImpl.java:138)
>>      at 
>> org.apache.http.impl.nio.DefaultNHttpClientConnection.consumeInput(DefaultNHttpClientConnection.java:266)
>>      at 
>> org.apache.http.impl.nio.DefaultHttpClientIODispatch.onInputReady(DefaultHttpClientIODispatch.java:165)
>>      at 
>> org.apache.http.impl.nio.DefaultHttpClientIODispatch.onInputReady(DefaultHttpClientIODispatch.java:51)
>>      at 
>> org.apache.http.impl.nio.reactor.AbstractIODispatch.inputReady(AbstractIODispatch.java:113)
>>      at 
>> org.apache.http.impl.nio.reactor.BaseIOReactor.readable(BaseIOReactor.java:159)
>>      at 
>> org.apache.http.impl.nio.reactor.AbstractIOReactor.processEvent(AbstractIOReactor.java:338)
>>      at 
>> org.apache.http.impl.nio.reactor.AbstractIOReactor.processEvents(AbstractIOReactor.java:316)
>>      at 
>> org.apache.http.impl.nio.reactor.AbstractIOReactor.execute(AbstractIOReactor.java:277)
>>      at 
>> org.apache.http.impl.nio.reactor.BaseIOReactor.execute(BaseIOReactor.java:105)
>>      at 
>> org.apache.http.impl.nio.reactor.AbstractMultiworkerIOReactor$Worker.run(AbstractMultiworkerIOReactor.java:584)
>>      at java.lang.Thread.run(Thread.java:722)
>> 
> 
> Yes, that is the price for shutting down a connection mid-air (so to
> speak). 
> 

Makes sense, yeah!

>> Is there a better way to do this safely?
>> 
> 
> You might just configure I/O reactor to silently ignore
> CancelledKeyException.

How can I do that? Am I right that I’d need to go pretty deep into the 
IOReactor thing to do that?

> A more complex, but likely more proper way would
> be to create a custom ConnectionReuseStrategy that is aware of unwanted
> connection routes or unwanted connection instances.
> ConnectionReuseStrategy can get hold of the underlying connection object
> from the HttpContext instance passed in as a parameter.
> 

Do you have an example on that? That sounds like a good approach but I guess I 
need a starting point :)


> Oleg
> 
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]

Reply via email to