Re: Slowdown on high-load machines with 3000 sockets

[Christian Schmid <webmaster@rapidforum.com>]

[-- Attachment #1: Type: text/plain, Size: 1178 bytes --]

Attached is another traffic-image. This one is with 2.6.10 and a 3/1 split, preemtive kernel, so all 
defaults.

The first part is where I throttled the whole thing to 100 MBit in order to build up a traffic-jam ;)

When I released it, it jumped up immediately but suddenly it goes down (each pixel is one second) 
Playing around with min_free_kbytes didnt help. Where it goes up again I set lower_zone_protection 
to 1024000 and where it goes down I set it to 0 again and where it goes up the last time... guess..

Well I was only able to test this with 3500 sockets. I will try to jam up tomorrow to reach 5000 
sockets to see if that is really the reason.

Nick Piggin wrote:
> Christian Schmid wrote:
> 
>> This issue has been tracked down more. This bug does NOT appear if I 
>> disable preemtive kernel.
>> Maybe this helps.
>>
> 
> Yes, it may help - can you boot with profile=schedule and get
> the results for say, a 30 second period while the application
> is experiencing problems?
> 
> So:
> 
> start application
> wait till it hits slowdown
> readprofile -r ; sleep 30 ; readprofile > schedprof.out
> 
> and send schedprof.out, please?
> 
> Thanks,
> Nick
> 
> 
> 

[-- Attachment #2: traffic2.png --]
[-- Type: image/png, Size: 2504 bytes --]