Re: serious performance regression in .41

[christos%astron.com@localhost (Christos Zoulas)]

In article <0E8BFCFC-C4CD-4CE3-8BAC-22EB8EBD3471%eis.cs.tu-bs.de@localhost>,
J. Hannken-Illjes <hannken%eis.cs.tu-bs.de@localhost> wrote:
>On 22 May 2014, at 03:15, Christos Zoulas <christos%astron.com@localhost> 
>wrote:
>
>> In article <lljiob$ha3$1%ger.gmane.org@localhost>,
>> Christos Zoulas <christos%astron.com@localhost> wrote:
>> 
>>> Here's an ugly fix (but internally only ugly fix):
>>> 
>>> http://www.netbsd.org/~christos/mntvnode_lock.diff
>
>Only unmount will use vflush() so this will not help.

Yes, I said that in my last mail. But it does help umount; a lot.
I actually first noticed the problem when I was rebooting. umounting
all my filesystem was taking many minutes!

>> Although this improves the situation the machine still has large pauses.
>> - compile a kernel with lockdebug.
>
>No performance tests with LOCKDEBUG or DEBUG please.  These options
>are known to be very expensive.

LOCKDEBUG and DEBUG were usable before. Now they are not. They are just
mentioned here to easily demonstrate the problem.

>> - start a build
>> - hold a key down at the shell prompt and notice how it behaves
>> 
>> Every 3 or so lines (of 80 characters), you'll notice a large pause
>> where the machine is unresponsive for a couple of seconds, and then
>comes back.
>
>I'm not able to reproduce this behaviour here.  Please describe
>your setup (mounted file systems and build command) in more detail.

The build command is standard. I have raid over sd0 (SAS drives).
How much memory do you have? Leave enough time for the buffer cache
to fill up.

>Will "systat vmstat 1" show anything during the pause?

The disk accesses stop completely and the machine pauses frequently.

christos