Setup Guide
You can enable the capability to detect uneven trac distribution in the member links of a HiGig link bundle on a line-card or SFM NPU. You
can also enable a notication to be sent using alarms and SNMP traps. The algorithm used to determine uneven distribution of trac is
predened.
Monitoring HiGig link bundles allows you to view and analyze unequal trac ow in backplane port channels and take corrective action.
Alarms are generated if the link-bundle trac threshold is greater than the congured threshold and the unevenness is greater than 10
percent between links for three successive rate-intervals. Alarms are removed when the link-bundle threshold is lower than the congured
threshold and the unevenness is less than 10 percent between links for three successive rate intervals.
An alarm includes the following information:
• Line-card or SFM NPU unit and HiGig port-channel ID in the format: hg-port-channel slot slot/npu-id/hg-port—
channel-id
• Alarm: triggered or cleared
Examples of the system log messages triggered when the threshold for a HiGig link bundle/port channel is exceeded are:
• %STKUNIT0-M:CP %SWMGR-5-HG-BUNDLE_UNEVEN_DISTRIBUTION: Found uneven distribution in hg-port-channel 0/5/0
• %STKUNIT0-M:CP %SWMGR-5-HG-BUNDLE_UNEVEN_DISTRIBUTION_ALARM_CLEAR: Uneven distribution in hg-port-
channel 0/5/0 got cleared
Guidelines for Monitoring HiGig Link-Bundles
When conguring HiGig link-bundle monitoring on the backplane, follow these guidelines:
• By default, the capability to monitor the trac distribution in a HiGig link bundle on a line-card or SFM NPU is disabled.
• Each line-card NPU uses two HiGig link bundle for its backplane links to connect each SFM (spine) NPU. The convention used to
identify a HiGig link-bundle interface is: hg-port-channel slot/npu-id/0, where slot species the line-card slot number (0–11), npu-id
species the NPU ID number (0), and HiGig port-channel ID which is in the range (0-2) for a line-card NPU
• Each SFM NPU uses a separate HiGig link bundle to connect to each line-card (leaf) NPU. The convention used to identify a HiGig link-
bundle interface is: hg-port-channel slot /higig-port-channel-id, where slot species SFM slot number which is in the range of 0 to 1,
npu-id species the NPU ID number (0)
• HiGig link-bundle monitoring starts only when:
– You enable monitoring for a specied HiGig link bundle using the hg-link-bundle monitor command.
– Bundle usage for egress trac exceeds the threshold congured with the hg-link-bundle monitor trigger-threshold
command.
Alarms are generated only when link-bundle trac levels are high. At low trac levels, only one or two signicant ows may cause
unevenness. However, uneven trac distribution across links during low-trac periods is not critical and does not trigger an alarm.
• You can enable SNMP traps and syslog messages to be generated when an uneven trac distribution is detected in a HiGig link bundle.
• Trac distribution in a HiGig link bundle is calculated as the bandwidth-weighted mean use of all links in the bundle. This calculation is
performed only on links that are up in their operational status.
• The rate interval used to poll trac distribution in member links in a HiGig link bundle is user-congurable. The default polling interval is
15 seconds.
• The trigger threshold species the percentage of total bundle bandwidth used to issue an alarm for uneven trac distribution. The
default is 60 percent. When the mean link utilization is below this value, uneven link-bundle trac is not reported.
The dierence in utilization percentage between the high-used link and low-used link determines the alarm condition. Alarm reporting
for link-bundle monitoring is based on the same algorithm used for LAG/ECMP. An alarm condition occurs when the unevenness in link-
bundle utilization exceeds 10 percent of the congured threshold and remains active until trac on member links falls below the trigger
Interfaces
471