AIX 5.2 stops forwarding traffic under load

[Phillip L. Kazakoff]

Hi guys,

We are testing an AIX 5.2 machine (p610) for use as a network router.
Under heavy load (the system is close to 100% CPU utilization) the
machine periodically stops forwarding traffic for a few seconds (CPU
utilization goes to zero) and then everything gets back to normal.

Does anyone has any idea as to what's wrong with the box and how to
fix it?

Tnanks

Phillip L. Kazakoff
Ramat-Gan, Israel

[Mark Taylor]

routing issue could be a by product of your cpu bound situation.



what processes are spinning at 100% ?



use ps aux, tprof, pprof, topas, nmon etc to work this out, then contact
the vendor of the application to which the processes belong to find out
why thier app is killing your system....



If its a kproc, then get the lastest ML installed to see if it resolves
the problem and if not, then collect some perfpmr data and open a
support call with IBM to see if the defect has been seen elsewhere or if
its a new defect .



if its a looping ksh script .. then re-write it in perl / python or "c"



If you can't control the app for some reason, then look at restricting
its cpu time with "wlm" - (i hate wlm, but if thats you only option,
then at least its an option)



Rgds

Mark


--
Posted via [url]http://dbforums.com[/url]

[Steve Nottingham]

[email]pkazakoff@bezeqint.net[/email] (Phillip L. Kazakoff) wrote in message news:<3c43fc0a.0309170116.478ce2f@posting.google.c om>...

> Hi guys,
>
> We are testing an AIX 5.2 machine (p610) for use as a network router.
> Under heavy load (the system is close to 100% CPU utilization) the
> machine periodically stops forwarding traffic for a few seconds (CPU
> utilization goes to zero) and then everything gets back to normal.
>
> Does anyone has any idea as to what's wrong with the box and how to
> fix it?
>
> Tnanks
>
> Phillip L. Kazakoff
> Ramat-Gan, Israel

1. Stop putting heavy load on it :-)
2. The CPU isn't powerful enough to do the work.
3. Buy a dedicated router e.g. Cisco.

Steve

[Phillip L. Kazakoff]

Problem is, at the time the problem occurs, i.e. the machine stops
forwarding traffic the CPU utilization is ZERO. The systems just does
not forward packets (just like the IP forwarding gets turned off).
When everything works, the CPU is utilized by the kernel (doing IP
forwarding).

Phill

Mark Taylor <member20596@dbforums.com> wrote in message news:<3379906.1063792962@dbforums.com>...

> routing issue could be a by product of your cpu bound situation.
>
>
>
> what processes are spinning at 100% ?
>
>
>
> use ps aux, tprof, pprof, topas, nmon etc to work this out, then contact
> the vendor of the application to which the processes belong to find out
> why thier app is killing your system....
>
>
>
> If its a kproc, then get the lastest ML installed to see if it resolves
> the problem and if not, then collect some perfpmr data and open a
> support call with IBM to see if the defect has been seen elsewhere or if
> its a new defect .
>
>
>
> if its a looping ksh script .. then re-write it in perl / python or "c"
>
>
>
> If you can't control the app for some reason, then look at restricting
> its cpu time with "wlm" - (i hate wlm, but if thats you only option,
> then at least its an option)
>
>
>
> Rgds
>
> Mark