Comments on: A Very Wonky Request/CFTHREAD Bug

By: Ken Auenson

Ken Auenson — Thu, 29 Nov 2007 02:32:32 +0000

Barney,
I ran this test on the vanilla install with the built-in web server on Windows XP… At least… I don't remember installing any of the patches or anything… too many machines to remember! lol

hope this helps, from CF Admin:
CF Version 8,0,0,176276
JVM 1.6.0_01

By: barneyb

barneyb — Thu, 29 Nov 2007 02:25:15 +0000

Ken,

Very interesting. Can I inquire about your environment (JVM, OS, etc.)? I'm running a JEE installation on Windows with Tomcat as the container.

By: Ken Auenson

Ken Auenson — Thu, 29 Nov 2007 02:20:02 +0000

Barney,
I spoke up way too soon after only reading your post without actually greping everything you were saying…
I get exactly what you are describing now!

I used your code and ran the tests on my dev edition, and I am not able to reproduce. If I run the first and second threads with the joins at the same time, they both run in seperate cfthread threads [different "Java Thread Name"s according to the server monitor]; and they do run in parallel [the second is not waiting for the first to complete].

HTH!

By: barneyb

barneyb — Thu, 29 Nov 2007 01:34:31 +0000

Ken,

It's not a limit on concurrent requests (I have mine set at ten). I can reach that by simply refreshing my three tabs four times each (for a total of 12 requests). Two threads queue up while ten run, and when the first two of the ten complete, the two queued ones start running.

With the joins, those requests (in aggregate) become single threaded for launching CFTHREADs, while other requests can happily run CFTHREADs concurrently.

By: barneyb

barneyb — Thu, 29 Nov 2007 01:32:15 +0000

Ken,

With Standard edition, you are limited in the number of threads, but if you've got that limitation, then the first run of three concurrent requests wouldn't load three threads, it'd load two and then one. So you'd still be able to differentiate it from both the correct and the observed behaviour.

By: Ken Auenson

Ken Auenson — Thu, 29 Nov 2007 01:30:14 +0000

Oops…
I guess I do need to get some sleep!
Since you do mention the server monitor in your post, this means you are on developer or enterprise edition, so please ignore my previous comment!
It sure does sound like there is a limiter on the simultaneous running threads, tho!

By: Ken Auenson

Ken Auenson — Thu, 29 Nov 2007 01:27:50 +0000

Barney,
Are you running these tests on CF8 standard?
I ask because CF8 standard does limit the number of threads that CFThread can use…
actually, the quote I am finding online is "CFTHREAD is limited to two additional spawned threads in Standard Edition"
That sounds like it could explain your observations, if some other process was already using one of the threads alotted to CFThread…

By: Ben Nadel

Ben Nadel — Wed, 28 Nov 2007 22:06:40 +0000

Insanity!

By: barneyb

barneyb — Wed, 28 Nov 2007 21:56:16 +0000

Ben,

I did try, actually, and it doesn't seem to matter.

By: Ben Nadel

Ben Nadel — Wed, 28 Nov 2007 21:49:50 +0000

That is very strange. Do you know if the same issue crops up if tab 1 and 2 are different files (ex. Test1.cfm, Test2.cfm). It might be that the threads are somehow tied to the physical file even though they have different requests?