Checksum Error Messages - Extreme Networks ExtremeWare Version 7.8 Troubleshooting Manual

Advanced system diagnostics

Table of Contents

Packet Errors and Packet Error Detection

The intent of these messages is to alert the NOC that the health check error threshold is being exceeded.

Closer monitoring is required, but these errors do not necessarily point to a systematic problem.

These messages take the general format:

date time < level : from > Sys-health-check type checksum error cat data

Example:

<CRIT:SYST> Sys-health-check [EXT] checksum error (fast path)

on slot 2,prev=73 cur=82

where:

level

The severity level, either CRIT or WARN.

•

CRIT indicates a critical failure in the system that requires corrective action.

•

WARN indicates that the error is most likely an isolated checksum occurrence and should

not be considered a critical failure to the system.

from

The source of the message, from either SYST (system) or KERN (kernel).

The type of test packet involved, possibly pointing to the packet origin and the diagnostic

type

subsystem involved. Supported packet type descriptors are: CPU, DIAG, EDP, EXT, and INT.

•

CPU, DIAG, and EDP refer to packets generated within the CPU health-checking

subsystem on the control (slow) path of the switch.

CPU packet types include DIAG and EDP packets, as well as data packets destined for

the CPU.

DIAG packet types refer to the CPU diagnostics packets generated by the health-checking

subsystem.

EDP packet types refer to the backplane health-checks used to test the communication

paths between the CPU and the I/O blades.

•

EXT and INT packet types refer to a checksum that is appended to each and every packet

that enters (ingresses) and exits (egresses) the switch fabric MACs on the data (fast) path.

EXT refers to an external fabric checksum (external, user-facing MAC); INT refers to an

internal fabric checksum (internal, backplane-facing MAC).

(Extreme " i " series switches pro-actively scan for fault conditions throughout the switch

architecture, and these packet types are all part of this effort. A checksum on one of these

packet types could have its root in packet memory, because all of these test packet types are

stored for a time in the packet memory. If the failure is within the packet memory and is

repeatable, run the packet memory scan to isolate and correct the failure.)

The category—slow path counters vs. fast path counters—associated with the health check

cat

error. This parameter indicates whether the window was marked bad due to fast-path reports

or slow-path reports.

Additional error information that summarizes the failure; content differs depending on the

data

associated type.