tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From André Warnier>
Subject Re: Working mod_jk related to loglevel with wildfly?
Date Tue, 10 Jun 2014 19:35:46 GMT
Mark Eggers wrote:
> Hash: SHA1
> On 6/10/2014 8:29 AM, André Warnier wrote:
>> André Warnier wrote:
>>> Martin Stolk wrote:
>>>> Hello,
>>>> We are migrating our applications from tomcat to wildfly. We 
>>>> are using mod_= jk (1.2.40) to connect apache to the wildfly 
>>>> ajp port.
>>>> When using tomcat there are no problems, but with wilfdly
>>>> there is a strang= e behavior in our application.
>>> It is a bit of a puzzle then, why you are asking for help here. 
>>> Would "" not be a better place to 
>>> start ?
>>>> Our application is written in java (wicket) and when entering a
>>>> search form= every field fills with a semi-colon after entering
>>>> the find button. When i= set the JkLogLevel to trace or debug
>>>> the problems remains but less frequen= tly and not in every
>>>> form. I also tried different ForwardURI** JkOptions, but that
>>>> make no difference.
>>> I can't think of a reason off-hand why this should ever make any
>>>  difference. It would seem that the first thing to look at, is 
>>> what this "Find" button in the form really does.  Is it just a 
>>> "submit" button, or does it call something (some javascript 
>>> perhaps) ? Does the <form> send a POST, or a GET request ?
>>>> Can anyone help me where to find a solution?
>> Ok, I'll bite again. As I understand the issue, you have the 
>> following schema :
>> B + BA <-HTTP-> A + M <-AJP-> E + EA
>> where :
>> - B is the browser - BA is the "application" in the browser. That 
>> can be pure HTML, or HTML + javascript, or a Java Applet, or 
>> whatever - A is the Apache httpd front-end - M is the mod_jk
>> module running inside Apache httpd - E is the Servlet Engine
>> (Tomcat or Wildfly) - EA is the java application running inside of
>> E
>> and we assume that the only element which varies is E, which is 
>> either Tomcat or Wildfly.
>> You say that when E is Tomcat, everything works fine. But when E
>> is Wildfly, strange things happen.
>> Given that B + BA are the same and would send the same HTTP 
>> requests in both cases to A, - there is no reason why A would do 
>> anything different when E is Wildfly, than when E is Tomcat. A
>> does not even know which Servlet Engine E is being used. - there is
>> no reason why M would do anything different when E is Wildfly,
>> than when E is Tomcat. M does not even know which Servlet Engine E
>> is being used. It just knows that it is talking to an AJP connector
>> of a webserver, and that it needs to "translate" the HTTP request,
>> to an AJP request, before forwarding it.
>> The only impact that I can think of, of changing the mod_jk 
>> loglevel, is to make mod_jk perhaps a little bit slower, because
>> it has to log more. (But we should be talking of at most
>> milliseconds here).
>> So, on the face of it, logically, I would think that if there is a
>>  problem when E is Wildfly, the problem must be with Wildfly, or 
>> with how Wildfly is running the EA application.
>> Or else, our premise is wrong, and BA is not exactly the same in 
>> both cases, and does not send exactly the same thing to A. But 
>> since BA "comes from" E + EA originally, that would also mean that
>>  the problem is with Wildfly + the EA application.
>> So I would still go to the Wildfly support list, present the same 
>> case as you did above, and ask them if they have a clue as to what 
>> may be happening.
> To extend André's excellent examination . . . .
> It would be nice if you could remove A + M from the equation. In other
> words:
> B + BA <-HTTP-> E + EA
> Then vary E (Wildfly or Tomcat).
> If both work, then the issue might be with Firefly's AJP configuration
> (or its AJP implementation).
> If Firefly does not work, then the issue might be with Firefly's
> configuration (or Firefly and Wicket).
> If neither work, then that's a puzzle.
> . . . . just my (coffee-less) 2 cents

Now wait, Firefly ?
Is that linked to the coffee-less state ?

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message