FC4 possible kernel 2.6.15 issue

[Skunk Worx]

Skunk Worx wrote:
> Skunk Worx wrote:
>> Skunk Worx wrote:
>>> I use the GSOAP soap server toolkit for transferring data.
>>>
>>> Prior to kernel 2.6.15 I could pull about 3000 data items a second 
>>> using  the various persistent http connection toolkit options over 
>>> TCP/IP.
>>>
>>> Under 2.6.15 I am lucky to get 100 items/sec using the same codebase.
>>>
>>
We're slowly tracing this through, it looks like the problem only 
appears through the lo interface (client and server on same machine).

Two 2.6.15 machines over ether don't exhibit the problem.

On 2.6.14, almost every strace'd recv() call is like this :

15:27:04.568800 recv(3, "<?xml version=\"1.0\" encoding=\"UT"..., 555, 
0) = 555 <0.000121>

...the call is a fraction of a millisecond.

Under localhost and 2.6.15, it's often around 40 ms.:

15:18:24.515891 recv(3, "<?xml version=\"1.0\" encoding=\"UT"..., 566, 0) =
566 <0.038414>

Under 2.6.14, there was one or two of these 40 ms. calls out of 48,000 
calls total. Under 2.6.15, we're seeing about 3,000 40 ms. of these 
calls out of 48,000 calls total.

I've also verified the problem using the stock kernel.org releases on my 
FC5T3 box.

We might be getting closer to the issue. Right now we're trying to 
decide whether or not to write a simple client/server pair of apps that 
demonstrate the problem.

--
SW