> This phenomenon appears to be caused by some interaction between the
> delivery of I/O interrupts and timer interrupts when running under
> VM (or LPAR with shared CPUs, for that matter). We haven't quite
> understood this whole issue yet, unfortunately. (Maybe Martin knows
> more details ...)
Ulrich Weigand wrote:
This once-a-second wakeup for kswapd looks like an artifact of the
2.4.7-period MM design, even 2.4.17 doesn't have that any more as
far as I can see.
My observations were indeed with the only kernel that currently is supported
for IBM middleware applications... I the outloo
Rob can der Heij wrote:
>Right now my biggest problem is kswapd that wants a wakeup call every
>second, and I would be interesting to hear if someone can explain why it
>is a smart way to implement memory management like this. If no work was
>done since the previous call of kswapd, it should be po
Ulrich Weigand wrote:
On the other hand I'm not sure that even with the timer patch and
without network connections you can get a Linux guest to go 'idle
enough' to drop completely out of queue; there's user space daemons
that get periodically active and the like.
Aber sicher! Right now my bigge
tes this E-Mail to Master Yoda )
> -Original Message-
> From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf
Of
> Rob van der Heij
> Sent: Monday, February 10, 2003 4:18 PM
> To: [EMAIL PROTECTED]
> Subject: Re: [LINUX-390] Timer patch on 2.4.19
>
> >Or they
>Or they or Boeblingen can figure out how to get those drivers to allow the
We believe we know what is causing this, and VM development understand the
implications for some installations and the need to fix this. Actually,
during the last days of the residency we already were able to demonstrate
Mark,
> One other "symptom." When I "echo 1 > hz_timer", the status of
> ksoftirqd_CPU0 becomes "SWN." When I "echo 0 > hz_timer" it becomes "RWN."
Yes, that's the problem. As I recall, the issue is that even though
ksoftirqd_CPU0 is really active only very rarely, for some reason the
timer in
Friday, March 07, 2003 4:52 PM
To: [EMAIL PROTECTED]
Cc: [EMAIL PROTECTED]
Subject: Re: Timer patch on 2.4.19
Mark Post wrote:
-snip-
>One other possible symptom is that your system
>load never goes below 1.00. That's what I'm seeing on my 2.4.19 systems
>anyway.
This is a
Mark Post wrote:
>Or they or Boeblingen can figure out how to get those drivers to allow the
>guest to drop from queue.
>From what I understood about this issue it would appear that CP decides
a guest isn't 'really idle' if it has any CCW I/O currently in progress.
However, due to the way networ
Sent by: Linux Subject: Re: Timer patch on 2.4.19
on 390 Port
<[EMAIL PROTECTED]
arist.edu>
03/07/2003
07:26 PM
Please respond
on [mailto:[EMAIL PROTECTED]
Sent: Friday, March 07, 2003 2:11 PM
To: [EMAIL PROTECTED]
Subject: Timer patch on 2.4.19
One thing Vic, I learned from Rob a couple weeks ago.
For some reason, servers using QDIO, CTC, etc (Other than
IUCV).do not drop from queue, thus go into queue 3, thus
do not
One thing Vic, I learned from Rob a couple weeks ago.
For some reason, servers using QDIO, CTC, etc (Other than
IUCV).do not drop from queue, thus go into queue 3, thus
do not get their storage trimmed. This is of course badness.
As Rich pointed out, to see if the timer patch is working,
look at t
I have two machines running the 2.4.19 SLES 8 kernel. When I put "0" into
hz_timer, I can see the utilization drop off the radar. When set to "1", it
is at the usual .3% (on a MP3000, G5).
According to ESAMON, the machines sit in Q3 58% of the time.
On Thursday 06 March 2003 08:34 pm, you wrote
March 06, 2003 9:34 PM
To: [EMAIL PROTECTED]
Subject: Timer patch on 2.4.19
G'day all,
Just went through the process of building a new 2.4.19 kernel with timer
patch
applied (linux-2.4.19-timer-1-may2002 from DevWorks). I wonder if anyone
has
seen the /proc/sys/kernel/hz_timer pseudofil
G'day all,
Just went through the process of building a new 2.4.19 kernel with timer patch
applied (linux-2.4.19-timer-1-may2002 from DevWorks). I wonder if anyone has
seen the /proc/sys/kernel/hz_timer pseudofile, and can describe its function.
Full story, read on...
When building the kernel, I
15 matches
Mail list logo
