Mladen Turk schrieb:
> [EMAIL PROTECTED] wrote:
>>              ((rc = ajp_connection_tcp_send_message(ae, op->request,
>> l)) != JK_TRUE)) {
>>              if (rc != JK_FATAL_ERROR) {
>> -                jk_log(l, JK_LOG_INFO,
>> +                jk_log(l, JK_LOG_DEBUG,
>>                         "(%s) error sending request. Will try another
>> pooled connection",
> 
> -1. Please change back to INFO because this is *not* part of
> the normal operation.
> If someone hates those messages he can increase the log level to warn.

That's true. But the message shows up, e.g. when the backend closed the
connection, and mod_jk just realized that via jk_is_socket_connected().
But OK, I'll revert.

>> @@ -1219,8 +1219,8 @@
>>              jk_log(l, JK_LOG_INFO,
>>                     "(%s) all endpoints are disconnected or dead",
>>                     ae->worker->name);
>> -            JK_TRACE_EXIT(l);
>> -            return JK_FALSE;
>> +            jk_log(l, JK_LOG_INFO,
>> +                   "Increase the backend idle connection timeout or
>> the connection_pool_minsize");
> 
> I'm not sure about that either.
> This is where retries should go into the action and be used for sticky
> sessions.
> If the sticky sessions are false, there might be other workers in the lb
> that
> are not disconnected.
> OTOH if the retries is larger then 1 (2 is default) the same worker will be
> used and the new connection will be made.

Yes, but in this case I think it's not really a retry. Assume prefork
model and the remote backend closed the connection. We will come very
often here and we should at least try once to open a new connection - as
we do after restarting the web server. In my opinion, this is not yet a
retry situation. I understand your argument concerning using a working
backend instead of connecting again, but we are far from that even with
the previous code. If maintenance closes the connections, then we will
nevertheless open new ones. This fragment here only comes into play, if
the backend closed the connections (timeout or failure). So I would
prefer keeping this change.

> Regards,
> Mladen.

Regards,

Rainer


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

Reply via email to