How to Use Net X-ray PC tool when reporting Network Faults at Night

Gary Mitchell 03-Sep-1998

Index

Introduction
Where are the PCs with the tool?
Is there a problem on the Network?
When there is congestion or excessive error rate

Introduction

These notes are intended to be used as a simple recipe book to identify the source of network congestion by Telescope Operators or Duty Engineers during the night.

return to index

Where are the PCs with the tool?

WHT: You should use the CFG PC located in the operations room and labelled "lplx02".
INT: You should use the PC located in the Clip Center room and labelled "????".

return to index

Is there a problem on the Network?

Always begin by determing if there is

congestion
excessive error rate

Unless one or other or both conditions are true there is no point in spending time gazing at traffic maps and reporting "well host x was exchanging a lot of packets with host y ". There is often busy traffic. The question is only "is the network coping?"

The most useful tool to start with is always " Dashboard"

From the top bar of Net-Xray select Tools --> Dashboard.

In the Dashboard window you can have "Gauge" or "Detail" mode. Set "Gauge" mode if necessary. Are any needles in the red zone? Are any needles near the red zone? Record average and peek levels of all 3 dials over about 1 whole minute.

return to index

When there is congestion or excessive error rate

Congestion: an average of 50% or peeking above 50% several times a minute.
Excessive error rate: an average of 1 per second or peeking above 1 per second several times a minute.

To identify contributors to congestion Bring up the "Matrix" tool.

From the top bar of Net-Xray select Tools --> Matrix.

It is not unusual for this diagram (a circle with chords) to get very complicated and with illegible labels. When this happens close the "matrix" window and re-start the matrix tool. At the bottom there a 3 modes

Set in IP mode. Identify the principle pairs contributing to the network congestion. The easiest way is to place the pointer on the thickest green chord and read the text window which pops up to describe the chord. There will usually be one or two significant pairs only when some task is overloading the network. The IP mode will not show up DECnet traffic between VAXes. If you don't see any obvious pair with Traffic Monitor in IP mode then switch to MAC.

Note that when congestion is high then this often leads to a high error rate because of many garbled packets due to collisions (simultaneous transmissions). If there are high error rates even though there is no congestion then this could be due to

physical damage to a network segment
a defective tranceiver
two hosts having the same IP address due to some misconfiguration.

return to index

Report what you found

Please check you have all the following items as in the example text below.

Examined network at HH:MM 
Dashboard averages over a minute were
Packets/s 100, Utilisation 40%, Errors 4%
Dashboard peeks over a minute were
Packets/s 1K, Utilisation 60%, Errors 16%
The traffic map had as the biggest chord
lpss15  - orion

return to index