CentOS - Apr 2016 - Kernel 2.6.32-573.22.1.el6.x86_64, higher than usual load

Hi,

After switching to the 2.6.32-573.22.1.el6.x86_64 kernel (CentOS 6) a few weeks
ago I have noticed that my servers are now reporting a much higher idle load.

The servers are being monitored with Zabbix, and there is a clear difference
between the older kernel 2.6.32-573.12.1 and the newer 2.6.32-573.12.1.

I can see the problem on both physical and virtual servers.

You can see the Load Average chart here from one of our servers:
http://imgur.com/U1eihMX

The server is pretty much idle all the time, but the load average increased from
0.001 to 0.300 when updated to the new kernel.

Is anyone else seeing the same behavior?

Best Regards,
Johnny Carlsen
________________________________
CONFIDENTIAL AND PRIVILEGED: This e-mail and associated content is subject to
the disclaimer statement found at http://aware.co.th/emaildisclaimer/rev1-0.php.

On 04/08/2016 06:50 AM, Johnny Carlsen wrote:
> Hi,
> 
> After switching to the 2.6.32-573.22.1.el6.x86_64 kernel (CentOS 6) a few
weeks ago I have noticed that my servers are now reporting a much higher idle
load.
> 
> The servers are being monitored with Zabbix, and there is a clear
difference between the older kernel 2.6.32-573.12.1 and the newer
2.6.32-573.12.1.
> 
> I can see the problem on both physical and virtual servers.
> 
> You can see the Load Average chart here from one of our servers:
> http://imgur.com/U1eihMX
> 
> The server is pretty much idle all the time, but the load average increased
from 0.001 to 0.300 when updated to the new kernel.
> 
> Is anyone else seeing the same behavior?
This is one of the fixes listed on the errata page:

"Due to prematurely decremented calc_load_task, the calculated load
average was off by up to the number of CPUs in the machine. As a
consequence, job scheduling worked improperly causing a drop in the
system performance. This update keeps the delta of the CPU going into
NO_HZ idle separately, and folds the pending idle delta into the global
active countwhile correctly aging the averages for the idle-duration
when leaving NO_HZ mode. Now, job scheduling works correctly, ensuring
balanced CPU load. (BZ#1300349)"

From:
https://rhn.redhat.com/errata/RHSA-2016-0494.html

No idea if that is impacting the load shown on your machine .. maybe
someone has some ideas.



-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: OpenPGP digital signature
URL:
<http://lists.centos.org/pipermail/centos/attachments/20160408/8588e087/attachment-0001.sig>

On Fri, Apr 8, 2016 at 6:32 AM, Johnny Hughes <johnny at centos.org>
wrote:
> On 04/08/2016 06:50 AM, Johnny Carlsen wrote:
> > Hi,
> >
> > After switching to the 2.6.32-573.22.1.el6.x86_64 kernel (CentOS 6) a
> few weeks ago I have noticed that my servers are now reporting a much
> higher idle load.
> >
> > The servers are being monitored with Zabbix, and there is a clear
> difference between the older kernel 2.6.32-573.12.1 and the newer
> 2.6.32-573.12.1.
> >
> > I can see the problem on both physical and virtual servers.
> >
> > You can see the Load Average chart here from one of our servers:
> > http://imgur.com/U1eihMX
> >
> > The server is pretty much idle all the time, but the load average
> increased from 0.001 to 0.300 when updated to the new kernel.
> >
> > Is anyone else seeing the same behavior?
>
> This is one of the fixes listed on the errata page:
>
> "Due to prematurely decremented calc_load_task, the calculated load
> (snip)"
>
> From:
> https://rhn.redhat.com/errata/RHSA-2016-0494.html
>
> No idea if that is impacting the load shown on your machine .. maybe
> someone has some ideas.
>
An RHEL user is reporting what seems to be the same issue:

https://access.redhat.com/discussions/2247501

Akemi