RE: cvs commit: jakarta-tomcat-connectors/jk/native2/common jk_channel_socket.c

[Mladen Turk]

> -----Original Message-----
> From: Ignacio J. Ortega [mailto:[EMAIL PROTECTED]] 
 
> 
> The comment is wrong, the problem is with the loadbalancer 
> logic, if one of the configured instances got lost, connect 
> returns WSAECONNREFUSED, with Malden's patch, the thread gots 
> stuck in a loop trying to connectto a non existing tc instance.. 
> 

Like I said you don't know if its non existing or just refusing more
connections due to maximum connections reached.
So what we need here is timout check inside connection phase, and not
giving up on the first failed connection request.
On the high load this will give you lots of 404's that are simply caused
by the fact that the TC is too busy refusing more concurent connections.

> Incidentally i have this config to test tc4 in port 8019 and 
> tc33 in 8009, i normally dont have the 2 versions working at 
> the same time, so i continually get WSAECONNREFUSED.. hence 
> the problem.. 
> 

If we put the timeout check inside the connection phase of the socket,
after that time we could mark the worker as disabled, or skip the
timeout testing for this worker, failing on the first connection error.
If the connection gets valid again, then we can check the timeout over
again.

MT.


--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>