Fabric OS Administrator’s Guide 419
53-1001763-02
Enabling bottleneck detection on a switch 18

Enabling bottleneck detection on a switch

Bottleneck detection is enabled on a switch basis. It is recommended that you enable bottleneck
detection on every switch in the fabric. If you add additional switches, including logical switches, to
the fabric, be sure to enable bottleneck detection on those switches as well.
When you enable bottleneck detection on a switch, the feature is applied to all eligible ports on that
switch. If ineligible ports later become eligible or, in the case of a logical switch, if ports are moved
to the logical switch, bottleneck detection is automatically applied to those ports.
Enabling bottleneck detection enables both latency and congestion detection.
When you enable bottleneck detection, you also determine whether alerts are to be sent when the
bottleneck conditions at a port exceed a specified threshold. These settings apply to all ports in the
switch.
1. Connect to the switch and log in as admin.
2. Enter the bottleneckmon --enable command to enable bottleneck detection on all eligible
ports on the switch.
By default, alerts are not sent unless you specify the alert parameter; however, you can view a
history of bottleneck conditions for the port as described in “Displaying bottleneck statistics”
on page 422.
3. Repeat step 1 through step 2 on every switch in the fabric.
NOTE
Best practice is to use the default values for the cthresh (0.8), lthresh (0.1), time (300), and qtime
(300) parameters. If you change the time parameter, you should use a setting that is 300 or higher.
Example of enabling bottleneck detection
(Preferred use case) The following example enables bottleneck detection on the switch with alerts
using default values for threshold and time.
switch:admin> bottleneckmon --enable -alert
The following example enables bottleneck detection on the switch without alerts. Although alerts
are not delivered in bottleneck conditions, you can view the bottleneck history using the CLI.
switch:admin> bottleneckmon --enable

Excluding a port from bottleneck detection

When you exclude a port from bottleneck detection, no data is collected from the port and no alerts
are generated for the port. All statistics history for the port is discarded.
Alerting parameters for the port are preserved, so if you later include the port for bottleneck
detection, the alerting parameters are restored.
Per-port exclusions might be needed if, for example, a long-distance port is known to be a
bottleneck because of credit insufficiency. In general, however, per-port exclusions are not
recommended.
For trunking, if you exclude a slave port from bottleneck detection, the exclusion has no effect as
long as the port is a trunk slave. The exclusion takes effect only if the port becomes a trunk master
or leaves the trunk.