[Orca-users] network vs. process swap

[Sean O'Neill]

At 10:41 AM 12/19/2002 -0700, Bruce Shaw wrote:

>On one of the boxes, Interface errors/secd varies between 2 and 14,
>averaging 2.7.  The network gods checked the switch statistics and found
>nothing exciting.

Sounds like bad hardware or a Sun/switch negotiation issue maybe.  You said 
this machine is being forced to 100Mbit FD right ?  I've seen cases where 
even with this forced setting the Sun and switch its attached to still have 
negotiation problems relating to 100Mbit/10Mbit or FD/HD.  A machine 
sitting right next to it of the same chassie with same config works just 
fine.  Moving the machine to another switch could fix it.

Have you watched the status light for the switch port the Sun is plugged 
into?  It should stay green (on a Cisco anyway) - with the usual blinks for 
activity.  Do you see it change color?  On a Cisco, orange means the switch 
and the machine are in a negotiation state (or something like that).

Try verifying that the data you are seeing in Orca is true - which is more 
then likely is.  Try using netstat -i to figure out which interface is 
giving you problems (assuming you have more then one).

Do you have more then one interface on the machine ?  Have you thought 
about trying to move the interface settings for the current interface over 
to the other one and disabling the current interface ?  If you do this, and 
the problem goes away, then the current interface is more then likely bad.

Are your patches up to date ?

>It just seems really wierd that for both boxes that have this severity 8
>network, they also have high processes swapped.

The swap outs and the network issue are probably not related but then who 
knows ...

>Any chance we're plotting the wrong thing?

Maybe but not likely.  Verify the Orca data as I mentioned above.

Sean

>-----Original Message-----
>From: Sean O'Neill [mailto:sean at seanoneill.info]
>Sent: Thursday, December 19, 2002 10:06 AM
>To: Bruce Shaw; 'orca-users at orcaware.com'
>Subject: Re: [Orca-users] network vs. process swap
>
>
>At 08:33 AM 12/19/2002 -0700, Bruce Shaw wrote:
>
> >My daily system overview for some of my systems shows "network", a red
>line,
> >pinned at 4.0
>
>This could be for several things.  You have a half-duplex interface
>?  Collisions will cause your network system overview line to spike
>somewhat like this as well retransmits and duplicate packets.  You need to
>look at your network related graphs to see what else is showing up.
>
> >When I go to the individual graphs, there's nothing particularly
>interesting
> >going on with the network.
>
>Something is making Orca spike this.  Is your Orca page publicly accessible
>?
>
> >However, the "processes in Run Queue/Waiting/Swapped" is a sea of red.
>
>This means:
>
>- You are swapping now at some rate AND have processes already swapped out
>- You had a swap out event in the past
>
>Processes that swap out of memory stay swapped out until something wakes
>them up.  If you have processes swapped out that stay swapped out, that
>isn't a problem.  The problem is when you start swapping in and out and it
>adversely hits you performance overall.  I would start by reviewing this
>graph and see when the swap out event began.  Your free memory and page
>scanning graphs should show activity at this same time.  Then try to figure
>out what was going on at that time which may entail asking folks what they
>were doing at the time on that box.
>
>
>--
>........................................................
>......... ..- -. .. -..- .-. ..- .-.. . ... ............
>.-- .. -. -... .-.. --- .-- ... -.. .-. --- --- .-.. ...
>
>Sean O'Neill
>
>This communication is intended for the use of the recipient to which it is
>addressed, and may contain confidential, personal and or privileged
>information. Please contact us immediately if you are not the intended
>recipient of this communication, and do not copy, distribute, or take action
>relying on it. Any communication received in error, or subsequent reply,
>should be deleted or destroyed.

--
........................................................
......... ..- -. .. -..- .-. ..- .-.. . ... ............
.-- .. -. -... .-.. --- .-- ... -.. .-. --- --- .-.. ...

Sean O'Neill