Re: crappy pm3's

John G. Thompson (jgt10@livingston.com)
Fri, 02 May 1997 08:50:29 -0700

At 03:15 PM 5/1/97 -0500, Jonah Yokubaitis wrote:
>On Thu, 1 May 1997, John G. Thompson wrote:
>
>|At 09:58 AM 5/1/97 -0000, Marc Powell wrote:
>|>On 5/1/97 5:18 AM, Edward Henigin sent these words of wisdom:
>|
>|[snip]
>|
>|>> 3) extremely sluggish performance, sometimes. Pings on local
>|>>10/100 baseT switched ethernet will be in the 100-700ms range, when
normally
>|>
>|>This also is something that we have seen on a fairly regular basis. It
>|>does not appear to be related to heavy usage of the box as it occurs on
>|>boxes that get very little traffic. Resetting the box fixes the problem.
>|>If anyone has any thoughts on this...
>|
>|The first thing that occurs to me is to check the ether0 netstats (sho
>|nets). My rule of thumb is that a network is health if Ierrs, Oerrs and
>|Resets are zero or single digit and the ratio of collisions divided by
>|Opkts is less than 5%. Anything else indicates possible problems on the
>|ethernet.
>
>Environment:
>
>Fully switched 10baseT environment using Cisco Catalyst 1900 switches
>uplinking to a Cisco Catalyst5000 switch using Full Duplex 100baseTX.
>We have several Ascend MAX4004s and USR TC hubs on the same Cat1900
>switch that the PM3s are on.
>
>The Cat5000 uplinks to a cisco 7513 via multi-mode fiber into a
>100baseFX PA in a vip2-20 card.
>
>All cabling is cat5 and is cat5 certified. It does not appear to be an
>ethernet problem.

It isn't clear to me from the above how many machines on each collision
domain. Is it one machine per domain/port?

>Here are the ether stats on some of our PM3s that are having problems:
>
>pm1.austin> sh nets
>Name Ipkts Ierrs Opkts Oerrs Collis Resets Queue
>ether0 40497473 204 40103900 0 316392 2 0
>pm1.austin> ver
>Livingston PortMaster PM-3 ComOS 3.5c6
>System uptime is 9 days 3 hours 35 minutes

There is a problem with the above machine. The Ierrs of 204 indicate a
problem. The 2 resets, if not done by administrative resets by !root also
indicate a problem. This machine does not appear to be on its own
collision domain.

>pm2.austin> sh nets
>Name Ipkts Ierrs Opkts Oerrs Collis Resets Queue
>ether0 885613 0 900601 0 3945 0 0
>pm2.austin> ver
>Livingston PortMaster PM-3 ComOS 3.5.1b11
>System uptime is 6 hours 51 minutes

The above machine looks okay, collisions under 1 percent.

>pm3.austin> sh nets
>Name Ipkts Ierrs Opkts Oerrs Collis Resets Queue
>ether0 15841913 0 16451627 0 89414 0 0
>pm3.austin> ver
>Livingston PortMaster PM-3 ComOS 3.5c6
>System uptime is 8 days 3 hours 3 minutes

The above machine looks okay, collisions under 1 percent.

>isdn-1.austin> sh nets
>Name Ipkts Ierrs Opkts Oerrs Collis Resets Queue
>ether0 128011 0 45632 0 0 0 0
>isdn-1.austin> ver
>Livingston PortMaster PM-3 ComOS 3.5c6
>System uptime is 13 hours 33 minutes

The above machine is on an excellent network even given the short uptime.

>isdn-2.austin> sh nets
>Name Ipkts Ierrs Opkts Oerrs Collis Resets Queue
>ether0 859378 0 925061 0 2101 0 0
>isdn-2.austin> ver
>Livingston PortMaster PM-3 ComOS 3.5.1b11
>System uptime is 14 hours 1 minutes

The above machine looks okay, collisions under 1 percent.

I would be interested in the increase rate of collisions during the period
of sluggish performance.

I would also be interested in a comparison of ping times to all machines on
the network at ten same time along with a diagram or map of what is
connected where.

JGT
---------------------------------------------------------------------------
John G. Thompson Livingston Enterprises Inc. Phone: (800) 458-9966
JOAT(MON) 4464 Willow Road Fax: (510)737-2110
support@livingston.com Pleasanton, CA 94588 http://www.livingston.com/
---------------------------------------------------------------------------
******* The solution to any problem lies in its proper definition. *******