Cisco ISE - Line Card ISE Configuration Manual Download

Page: 83 / 102

Software Configuration of ATM ISE Line Cards for Cisco 12000 Series Routers

Troubleshooting ATM Errors on the 4-Port ATM ISE Line Card

Cisco IOS Release 12.0(27)S

Improper queueing configuration, using the queue-limit or random-detect commands, allows for long
queues to build up in the SAR. In this case, if some VCs are oversubscribed, their queues might
eventually exhaust the TX cell buffer pool. When no TX cell buffers are available, no VC can enqueue
outbound packets, including such VCs that are not oversubscribed; that is, VC isolation is not
maintained. This can be confirmed by checking the following counters while

tx_resource_err

incrementing:

•

the

tx_queue_depth

counter should be zero and

•

the

tx_random_drops

and

tx_thresh_drops

should be constant, not incrementing.

In a properly configured system, the

tx_resource_err

counter of a VC should only be incrementing when

this specific VC is oversubscribed and dropping packets because the queue limit is being exceeded or
because of WRED operation. This can be confirmed by checking the following counters while

tx_resource_err

is incrementing:

•

If queue limit is configured on the VC, the

tx_queue_depth

counter should be non-zero, up to the

configured queue limit at maximum.

•

If WRED is configured, then

tx_queue_depth

should be non-zero, and either

tx_random_drops

tx_thresh_drops

(or both) will also be incrementing.

Note

The above will be observable easily on a VC that is constantly oversubscribed, so that (some of) its
queues are never completely drained. However, if the traffic sent to the VC is highly bursty, then the
queues might be building up momentarily, but be completely drained by the time the counters are
displayed. In this case,

tx_resource_err

will increment, even though the queue depth might be small (or

zero) at the time it is displayed. It is therefore recommended to sample the counters several times before
drawing conclusions.

Upgrading the FPGA Image

If the line card does not boot and you receive an error message indicating that there is a problem with
the Field-Programmable Gate Array (FPGA) image (or if the line card alphanumeric LED display
remains frozen in IOS STRT state), you need to upgrade the FPGA image using the

update-fpga

option

in the

diag

command.

Note

The

diag

command and the

update-fpga

option are documented in the

Field Diagnostics for the Cisco

12000 Series Internet Router

publication.

When the Cisco IOS image boots, it verifies that a compatible FPGA image is running on the router. The
major version number of the FPGA image must be the same as that expected by the Cisco IOS image;
the minor version number on the FPGA image must be the same as or greater than the minor version
number expected by the Cisco IOS image. For example, if the Cisco IOS image expects a minimum
FPGA image of 03.02, the software will verify that the actual major version number of the FPGA image
in the line card bootflash is 03, and that the minor version number is 02 or above.