tomcat-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Filip Hanik - Dev Lists <>
Subject Re: 6.0.x request processing patch
Date Tue, 25 Sep 2007 20:25:27 GMT
Rainer Jung wrote:
> Filip Hanik - Dev Lists wrote:
>> Rainer Jung wrote:
>>> Hi Filip,
>>> I guess that also fixes the misleading html manager display (request 
>>> times going up and up). That would be very nice.
>>> Just a quick shot:
>>> - maybe you can fix Procssing -> Processing everywhere
>> not sure what you mean?
> Simply a typo. You used everywhere
>    "lastRequestProcssingTime"
> instead of
>    "lastRequestProcessingTime".
>                    ^
> -------------------|
roger that
>>> - will this change the behaviour of the requestProcessingTime as one 
>>> can retrieve from the MBeans of the GlobalRequestProcessor or 
>>> Servlets? I guess (and hope) no?
>> yes it does, requestProcessingTime is incorrect, has always been, as 
>> it goes up and up when the request is not in use. Why wouldn't you 
>> want to change it?
> As far as I know, the accumulated request processing times retrieved 
> via the MBeans for Servlets or the GlobalRequestProcessor are correct. 
> I simply don't want to break them.
> I had a quick look, and our confusion might come from the fact, that 
> RequestInfo contains a processingTime and a requestProcessingTime. The 
> first one gets updated when the request is recycled and accumulates 
> the processing times. That's the one that always worked. The other one 
> is e.g. used when showing a full status in the html manager and is the 
> one you want to fix.
> You might consider to return as requestProcessingTime instead of "0" 
> when no request is running (as you suggest) the time taken for the 
> last request before as detected by updateCounters(). Both ways have 
> their advantages (detecting idleness vs. detecting last request time). 
> I didn't check though, if updateCounters is called unconditionally 
> (wrt. configuration).
again, I am very confused. not sure why you are talking about 
processingTime, since that property has nothing to do with the patch, 
and is working.
requestProcessingTime, is the time spent in the current request, if 
there is no request active, then it returns 0
>>> - lastRequestProcssingTime: This item is new but apart from the new 
>>> setter and getter not used, especially not in the 
>>> getRequestProcessingTime fix. If this data is needed, shouldn't we 
>>> also expose it via the mbean descriptor? It's nice for a statistic 
>>> sampling of response times.
>> where is the mbean-descriptor for this? I thought JMX just pulled it 
>> out using reflection
> Indeed. I don't really understand, which method gets used when. The 
> session managers e.g. have the processingTime in the mbeans 
> descriptors, as well as standard context and standard wrapper. 
> GlobalRequestProcessor and Servlet MBeans seem to use reflection.
besides the typo, I'm unclear on the rest of your comments and how they 
pertain to the patch :)


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

View raw message