Troubleshooting; Resource Exhaustion; Classifier Block Formation - Avaya ERS 45 Series Technical Configuration Manual

4.4 Troubleshooting

Special consideration should be given to a few areas where issues are more commonly encountered
during traffic-profile filter set construction and installation.
4.4.1

Resource Exhaustion

Filter set installation consumes precedence levels/masks and filter/meter/counter resources. There is a
finite amount of these resources. Furthermore, these resources are shared by QoS and non-QoS
applications and may also shared across ports in a cluster. For these reasons, resource exhaustion is
one of the more common issues that are encountered. Troubleshooting resource exhaustion issues is
generally straightforward. Error messages are generated when the requested resources are not available
during filter set construction:
% Evaluation precedence conflict (Traffic Prof filter set data)
% Traffic Prof filter set rule count exceeds limit
% Traffic Prof filter set policy count exceeds limit
Resource-related error messages are generated during filter set installation as well:
% Traffic Prof filter set elem count exceeds available resources
% Traffic Prof filter set install operation failed
% Traffic Prof filter set update operation failed
Issues encountered during filter set construction are relatively easy to diagnose based on the provided
error message. Filter set installation issues require a bit more effort to understand. The first step is always
to perform an inventory of the filter set components to determine the precedence and filter/meter/counter
resources that are needed. Next the resources available on the target port need to be considered. Are the
required consecutive precedence levels available? Are there enough filter/meter/counter resources
available to support an instance of the filter set? The 'show qos diag' output can be used to answer
these questions.
The different types of QoS hardware complicate resource issue diagnosis. With legacy ERS stackable
platforms (ERS55XX), resources are allocated on a per-port basis. With the newer platforms
(ERS4XXX/ERS56XX/ERS59XX/VSP7000), resources are allocated from a central pool that serves a
collection of ports (referred to in this document as a "port cluster"). With a centralized resource scheme,
resources consumed on one port can impact resource availability for all other ports in the same cluster.
Once again, the 'show qos diag' output can be consulted for cluster-wide resource utilization information.
4.4.2

Classifier Block Formation

As with resource issue diagnosis, different types of QoS hardware complicate classifier block formation.
Classifier elements matching the exact same protocol fields or the exact same portions thereof (i.e., the
fields are masked with the same bit-mask) are always block-compatible. On the ERS stackable platforms
(i.e., ERS4XXX/56XX/59XX/VSP7000) these restrictions have been eased. Filter criteria matching the
same protocol fields, regardless of the whether full or partial field matches are specified, can be combined
into a block. Furthermore, classifier elements matching different but compatible fields (i.e., fields from a
hardware-defined set of compatible fields) may also be combined into a block. This last class of potential
block members, however, is sometimes difficult to determine.
The QoS hardware on the ERS stackable platforms (ERS4XXX/ERS56XX/ERS59XX/VSP7000) builds a
'key' which represents the classification criteria that is used to identify traffic. The same classification data
can be represented several different ways (i.e., different keys can be built to match the same protocol
fields). If the keys representing different classifiers are the same, the classifiers may be combined into a
September 2015
Avaya Inc. – External Distribution
avaya.com
28