You are on page 1of 16

Switch Fabric- Troubleshooting tips

How to Troubleshoot Switch Fabric?
Introduction
What is Switch Fabric?
The switch fabric is essentially the backplane for all ports and modules on the switch
module. When a connection is made from a port on one module to a port on another
module, it is made across the switch fabric. Physically, it is the combination of silicon,
plastic, and metal that enables ports to connect and pass traffic between themselves.
The switch fabric can be blocking or non-blocking. Non-blocking fabric ensures that the
total bandwidth of all ports that use the switch fabric do not exceed its capacity. What this
means is that the density of the ports on the switch are such that their total capacity will
never be greater than that of the switch fabric. Switches operating in non-blocking mode
ensure that congestion will never occur on the switch, nor will ports ever want for
bandwidth between each other.
A blocking switch has a port density capacity that exceeds the total capacity of the switch
fabric. Control is possible by blocking traffic flow when the switch fabric capacity is
exceeded or otherwise not available.
The switch fabric resides on the SE. When a port has to communicate with another port,
it has the supervisor check its tables (Content Addressable Memory [CAM] for Layer 2
addresses and Ternary CAM [TCAM] for Layer 3 addresses) to determine what slot and
port it needs. The supervisor then establishes the connection between the ports.The
switch fabric can also reside on its own module (such as the Switch Fabric Module 2
(WS-X6500-SFM2) and the Switch Fabric Module (WS-C6500-SFM for the Catalyst 6500
Series), which enables the available capacity to be expanded without replacing the SE,
or to expand beyond the capacity of the SE.

On Catalyst 6500, Switch fabric is a daughter card installed at Sup720 , it used to be a
separate module, Switch Fabric module in its first implementation back in the days of
Sup2. It is used to provide backplane connectivity between linecards. The default
bandwidth available on the backplane of 6500 is 32 Gbps. This 32 Gbps is used by all

1

slots for serial transmission of data. Therefore at any instant only two ports can be
communicating.
With the addition of Switch fabric, the switch’s backplane changes from serially-accessed
bus to crossbar fabric. By using crossbar fabric, many ports can be simultaneously
transmitting and receiving data, providing a much higher throughput.

The crossbar fabric consists of 18 fabric channels, providing each linecard two fabric
channels into the crossbar fabric. These channels can run at 8Gbps or 20Gbps
depending upon the line card used. The CEF256 and dCEF256 series modules connect
to fabric using 8 Gbps per channel and CEF720 series modules connect to it using 20
Gbps per channel.

Requirements
For a module to use switch fabric, it should be a fabric enabled module

Troubleshooting Tips
1. If the Fabric Switch Module does not work as expected, check the following:
2

a) Check if the Fabric Switch Status is Active. To do this, use the show fabric active
command. This command will display the current status of the Fabric Switch. Here is an
example.

Switch# show fabric active
Active fabric card in slot 5
No backup fabric card in the system
If the system has backup fabric card, then:
Switch #show fabric active
show fabric active:
Active fabric card in slot 5
Backup fabric card in slot 6

b) Check the fabric status of switching modules in the device. To do this use the show
fabric status [slot_number | all] command. This command will display the fabric status
of one or all switching modules. Here is an example,
Switch# show fabric status
slot

channel

speed

module

status

status

fabric

1

0

8G

OK

OK

5

0

8G

OK

Up- Timeout

6

0

20G

OK

Up- BufError

8

0

8G

OK

OK

8

1

8G

OK

OK

9

0

8G

Down- DDRsync

Switch#

3

OK

c) Check the fabric utilization of switching modules. To do this use the show fabric
utilization [slot_number | all] command. This command will display the fabric utilization
of one or all modules.
Here is an example,

Switch# show fabric utilization all
slot

channel

speed

Ingress %

1

0

20G

0

0

1

1

20G

0

0

2

0

20G

0

24

2

1

20G

0

24

3

0

20G

48

0

4

0

20G

0

0

4

1

20G

0

0

Egress %

2. In certain rare condi tions out put of 'show fabric channel-counters' may show
incrementing number of rxErrors.
Switch#show fabric channel-counters
Slot channel rxErrors txErrors

4

txDrops lbusDrops

1

1

0

0

0

0

3

0

0

0

0

0

3

1

0

0

0

0

5

0

5

0

0

0

8

0

39

0

0

0

8

1

0

0

0

0

a) RxRrror indicates that the module received corrupted packet(s) and dropped.
b) The Fabric do NOT check CRC when forwarding frames between different fabric
ports/channels.
c) This could be due to the receiving module corrupting the frames or receiving
corrupted frames from any fabric-enabled module in the switch.

The following actions can be taken to solve these errors:
a) Reseat the module with rxErrors. Reloading the linecard in question might stop the
errors for some time, but the errors might eventually come back.
b) If empty slot is available in the chassis move the affected line card to empty slot.
c) If no empty slots available, swap the linecard that counts rxErrors with other linecard
within the chassis (with no issue) or good known working linecard.
d) Swap the active and standby supervisors (i.e. move supervisor from slot 5 to slot 6
and vice versa. Sup failover.
e) Replace the affected linecard.
If the output of "show fabric status" command is showing "not-hot" for linecards under
hotStandby support.

Switch#show fabric status
slot channel speed module fabric hotStandby Standby Standby
status status
3
5

0 20G

OK

support module fabric
OK Y(not-hot)

3

1 20G

OK

OK Y(not-hot)

4

0 20G

OK

OK Y(not-hot)

4

1 20G

OK

OK Y(not-hot)

5

0 20G

OK

OK

N/A

5

1 20G

OK

OK

N/A

6

0 20G

OK

OK

N/A

6

1 20G

OK

OK

N/A

7

0 20G

OK

OK Y(not-hot)

7

1 20G

OK

OK Y(not-hot)

8

0 20G

OK

OK Y(not-hot)

8

1 20G

OK

OK Y(not-hot)

9

0 20G

OK

OK Y(not-hot)

9

1 20G

OK

OK Y(not-hot)

Reason: The standby fabric hot sync feature is only supported on the E version of the
6500 chassis, and this system has a non-E version.
3. If you see the error message as, “SP: Linecard endpoint of Channel 7 lost
Sync.
To Lower fabric and trying to recover now!”.
Reason: The message caused by a line card not being fully or properly seated. To
identify this line card - the capture of show fabric fpoe map command need to be
analyzed. Here is an example, 同上

6

Switch#show fabric fpoe map
slot channel fpoe
1

0

0

1

1

9

2

0

1

2

1

10

3

0

2

3

1

11

4

0

3

4

1

12

5

0

4

6

0

5

6

1

14

7

0

6

7

1

15

8

0

7

8

1

16

9

0

8

9

1

17

Workaround: The fpoe will be mapped to a specific line card slot. Once the suspect line
card is identified From the output of show fabric fpoe map, fpoe 7 points to the line
7

card in slot 8 and that is card that is causing the error messages.your next action
should be to schedule a removal and re-insert of that card to try to eliminate this
message from re-occurring.
4. If the system switching performance drops from 30Mpps to 15Mpps.
Reason: When classic and fabric enabled modules are mixed in a chassis, the system
switching performance drops from 30Mpps to 15Mpps.

Older "Classic" modules in the 6500, models 61xx, 62xx, 63xx, 64xx, send all traffic over
the switch BUS backplane, to be forwarded by the supervisor. Fabric enabled modules
only send the packet headers over the bus and the switch fabric can be utilized for
forwarding the data portion of the packet.
Workaround: Consider replacing any "Classic" modules with fabric enabled modules, in
order to increase system performance.
5. To troubleshoot further, collect the following show command output before
opening a TAC case.
step 1. turn on service internal.
switch# configure terminal
switch(config)# service internal

step 2. collect the requested logs.
terminal length 0
show fabric active
show fabric channel-counters
show fabric drop
show fabric errors
show fabric errors threshold
8

show fabric fpoe map
show fabric status
show fabric utilization
show tech-support
remote login switch
terminal length 0
show fabric error
show fabric state-machine channel state
show fabric state-machine channel event_trace 11
show fabric resync
show fabric timeout
show platform hardware capacity fabric
exit

step 3. turn off service internal

switch # configure terminal
switch(config)# no service internal

9

Troubleshooting Example
1. Fabric Time out Error Message:
%FABRIC-SP-[module-number]-TIMEOUT_ERR: Fabric in slot [dec] reported timeout
error for channel [dec] (Module [dec], fabric connection [dec])
Description
The error message indicates that firmware code on the fabric detected that the input or
output buffer was not moving. To recover from this condition, the system will
automatically
resynchronize the fabric channel.
Troubleshooting Steps
1. Issue the command “hw-module reset” to soft-reset the module. After the module
is up again,
2. capture the output of the command “show module” and the command “show
diagnostic module all”.
Sample Output Of “show module”

Show Module
Mod Ports

Card Type

Model

Serial No.

--- ----- -------------------------------------- ------------------ --------- -----------------------------------2

24

CEF720 24 port 1000mb SFP

WS-X6724-SFP

SAL0AAAAAAA

3

24

CEF720 24 port 1000mb SFP

WS-X6724-SFP

SAD0AAAAAAA

5

2

Supervisor Engine 720

(Hot)

WS-SUP720-3B

SAD0AAAAAAA

6

2

Supervisor Engine 720

(Active)

WS-SUP720-3B

SAD0AAAAAAA

10

7

4

CEF720 4 port 10-Gigabit Ethernet WS-X6704-10GE SAL1AAAAAAA

8

4

CEF720 4 port 10-Gigabit Ethernet WS-X6704-10GE SAL1AAAAAAA

Sample Output of “show diagnostic module all”
Switch#show diagnostic module all
Current bootup diagnostic level: minimal
Module 6: Supervisor Engine 720 (Active)
Overall Diagnostic Result for Module 6 : PASS
Diagnostic level at card bootup: minimal
Test results: (. = Pass, F = Fail, U = Untested)
1) TestScratchRegister -------------> .
2) TestSPRPInbandPing --------------> .
3) TestTransceiverIntegrity:
Port

1 2

---------U

U

4) TestActiveToStandbyLoopback:
Port

1 2

---------U

U

5) TestLoopback:

11

Port

1 2

--------6) TestNewIndexLearn ---------------> .
7) TestDontConditionalLearn --------> .
8) TestBadBpduTrap -----------------> .
9) TestMatchCapture ----------------> .
10) TestProtocolMatchChannel --------> .
11) TestFibDevices ------------------> .
12) TestIPv4FibShortcut -------------> .
13) TestL3Capture2 ------------------> .
14) TestIPv6FibShortcut -------------> .
15) TestMPLSFibShortcut -------------> .
16) TestNATFibShortcut --------------> .
17) TestAclPermit -------------------> .
18) TestAclDeny ---------------------> .
19) TestQoSTcam ---------------------> .
20) TestL3VlanMet -------------------> .
21) TestIngressSpan -----------------> .
22) TestEgressSpan ------------------> .
23) TestNetflowInlineRewrite:
Port

12

1 2

---------U

U

24) TestFabricSnakeForward ----------> .
25) TestFabricSnakeBackward ---------> .
26) TestTrafficStress ---------------> U
27) TestFibTcamSSRAM ----------------> U
28) TestAsicMemory ------------------> U
29) TestAclQosTcam ------------------> U
30) TestNetflowTcam -----------------> U
31) ScheduleSwitchover --------------> U
32) TestFirmwareDiagStatus ----------> .

In case the output doesn’t come as expected, physically pull out and reseat the module
firmly in the chassis to hard-reset the module. After the module is up again, capture the
output of the command “show module” and “show diagnostic module all”
Here is an example of failed diagnostic test for module 1
Module 1: Catalyst 6000 supervisor 2 (Active) SerialNo :
Overall Diagnostic Result for Module 1 : MINOR ERROR
Diagnostic level at card bootup: minimal
Test results: (. = Pass, F = Fail, U = Untested)
1) TestSPRPInbandPing --------------> F
2) TestTransceiverIntegrity:

13

2. Overruns on some ports on Card 5 (WS-X6548-GE-TX)
Switch1# show interface counters
Noticed "overruns" on 4 interfaces. They were not incrementing
GigabitEthernet5/1 is up, line protocol is up (connected)
Full-duplex, 1000Mb/s, media type is 10/100/1000BaseT
0 input errors, 0 CRC, 0 frame, 26 overrun, 0 ignored
GigabitEthernet5/6 is up, line protocol is up (connected)
Full-duplex, 1000Mb/s, media type is 10/100/1000BaseT
0 input errors, 0 CRC, 0 frame, 14 overrun, 0 ignored
GigabitEthernet5/9 is up, line protocol is up (connected)
Full-duplex, 1000Mb/s, media type is 10/100/1000BaseT
0 input errors, 0 CRC, 0 frame, 159 overrun, 0 ignored
GigabitEthernet5/12 is up, line protocol is up (connected)
Full-duplex, 1000Mb/s, media type is 10/100/1000BaseT
0 input errors, 0 CRC, 0 frame, 269 overrun, 0 ignored
Also will enabling switch fabric rectify the overruns?

Solution:
Enabling switch fabric will not rectify overruns because after the installation of a Switch
Fabric Module in a Cisco Catalyst 6500 series switch, the traffic is forwarded to and from
modules in different modes which doesn't necessarily facilitate resolution for overruns.
The traffic is forwarded in one of these of these modes:

14

a) Flow-through mode: In this mode, data passes between the local bus and the
supervisor engine bus. This mode is used for traffic to or from modules that are not
fabric-enabled.
b) Truncated mode: Only truncated data (the first 64 bytes of the frame) goes over the
switch fabric channel if both the destination and the source are fabric-enabled modules. If
either the source or destination is not a fabric-enabled module, the data goes through the
switch fabric channel and the data bus. The Switch Fabric Module does not get involved
when traffic is forwarded between modules that are not fabric-enabled.
c) Compact mode: A compact version of the DBus header is forwarded over the switch
fabric channel, which delivers the best possible switching rate. Modules that are not
fabric-enabled do not support the compact mode and generate cyclic redundancy check
(CRC) errors upon receipt of frames in compact mode. This mode is used only when no
such modules are installed in the chassis.
Let’s understand what overrun is:Overrun - The number of times the receiver hardware was unable to hand received data
to a hardware buffer
Common Cause - The input rate of traffic exceeded the ability of the receiver to handle
the data.
From the given example, the module used is WS-X6548-GE-TX:
This module is 8:1 oversubscribed. The ports on this module go to servers. On these
modules there is a single 1-Gigabit Ethernet uplink from the port ASIC that supports eight
ports. These cards share a 1 Mb buffer between a group of ports (1-8, 9-16, 17-24, 2532, 33-40, and 41-48) since each block of eight ports is 8:1 oversubscribed. The
aggregate throughput of each block of eight ports cannot exceed 1 Gbps. These line
cards are oversubscription cards that are designed to extend gigabit to the desktop and
might not be ideal for server farm connectivity. For more information refer to:-

Troubleshooting Switch Port and Interface Problems
To resolve this issue of overrun, move the high volume servers to ports on different asic
groups, so that the traffic flow through the 8 ports of every asic group does not exceed 1
Gbps. Alternatively look for other ideal design recommendations on line cards that have
better oversubscription ratio.

It refers from: https://supportforums.cisco.com/document/71201/switch-fabrictroubleshooting-tips
15

More related:
The Differences of XFP, SFP and SFP+
How to Identify the Genuine Cisco SFP?
Basic Knowledge on Cisco SFP Modules
Technical Specifications for Cisco SFP Modules

More Cisco products and Reviews you can visit: http://www.3anetwork.com/blog

3Anetwork.com is a world leading Cisco networking products supplier, we supply original
new Cisco networking equipments, including Cisco Catalyst switches, Cisco routers,
Cisco firewalls, Cisco wireless products, Cisco modules and interface cards products at
competitive price and ship to worldwide.
Our website: http://www.3anetwork.com
Telephone: +852-3069-7733
Email: info@3Anetwork.com
Address: 23/F Lucky Plaza, 315-321 Lockhart Road, Wanchai, Hongkong

16