Troubleshooting a marginal link – Dell POWEREDGE M1000E User Manual

Page 45

Advertising
background image

Fabric OS Troubleshooting and Diagnostics Guide

31

53-1001769-01

Marginal links

3

Only one frame is transmitted and received at any given time. An external cable is not required to
run this test. The port LEDs flicker green rapidly while the test is running.

Table 5

shows the different loopback modes you can use when using the portLoopbackTest to test

a marginal link.

Troubleshooting a marginal link

1. Enter the portErrShow command.

2. Determine whether there is a relatively high number of errors (such as CRC errors or ENC_OUT

errors), or if there are a steadily increasing number of errors to confirm a marginal link. Sample
the data every 5 minutes until you see the counters increment.

The frames tx and rx are the number of frames being transmitted and received.

The crc_err counter are frames with CRC errors. If this counter goes up then the physical
path should be inspected. Check the cables to and from the switch, patch panel, and other
devices. Check the SFP by swapping it with a known good working SFP.
If you see this issue on an 8 Gbps blade, use the portCfgFillWord command to reduce EMI.

The crc_g_eof counter are frames with CRC errors and a good EOF. The first port detecting
a CRC error marks the frame with a bad EOF and passes the frame on to its destination.
Subsequent ports in the path will also detect the CRC error and the crc_err counter will
increment on these ports. However, since the first port marked the frame with a bad EOF,
the good EOF counter on the subsequent ports will not increment. The marginal link
associated with the port with an increasing good EOF counter is the marginal link and the
source of the errors.

The enc_out are errors that occur outside the frame and usually indicating a bad primitive.
To determine if you are having a cable problem, take snapshots of the port errors by using
the portErrShow command in increments of 5 to 10 minutes. If you notice the crc_err
counter go up, you have a bad or damaged cable, or a bad or damaged device in the path.

NOTE

ICLs see enc_out errors when ports on one side of the link are disabled.

The disc_c3 errors are discarded class 3 errors, which means that the switch is holding
onto the frame longer than the hold time allows. One problem this could be related to is
ISL oversubscription.

TABLE 5

Loopback modes

Loopback mode

Description

1

Port Loopback (loopback plugs)

2

External Serializer/Deserializer (SerDes) loopback

5

Internal (parallel) loopback (indicates no external equipment)

7

Back-end bypass and port loopback

8

Back-end bypass and SerDes loopback

9

Back-end bypass and internal loopback

Advertising