Guidelinesandlimitationsforcongestiondetection,Avoidance, And Isolation; Guidelines And Limitations For Congestion Detection - Cisco MDS 9000 Series Configuration Manual

Congestion Detection, Avoidance, and Isolation
GuidelinesandLimitationsforCongestionDetection,Avoidance,
and Isolation

Guidelines and Limitations for Congestion Detection

The show tech-support slowdrain command contains all the congestion detection indications, counters, and
log messages as well as other commands that allow an understanding of the switches, MDS NX-OS versions,
and topology. Since, congestion can propagate from one switch to another, the show tech-support slowdrain
command should be gathered from all the switches at approximately the same time to have the best view of
where the congestion started and how it spread. This can be easily done via the DCNM SAN client using the
Tools-> Run CLI feature. This feature will issue a command or commands to all the switches in the fabric
and consolidates the individual switch outputs into a single fabric zip file.
Some commands display simple counters such as the show interface counters command, whereas some
commands display counter information with accompanying date and time stamps. The commands that display
counters with accompanying date and time stamps are mostly the show logging onboard commands.
There are various "sections" of show logging onboard that contain information pertaining to slow drain and
over utilization. Most "sections" will update periodically and include counters only when they actually change
in the prior interval. Different sections have different update periods. They are:
• Error-stats—Includes many error counters accompanying date and time stamps
• Txwait—Includes interfaces that record 100ms or more of TxWait in a 20 second interval. The values
• Rxwait—Includes interfaces that record 100ms or more of RxWait in a 20 second interval. The values
When a counter increments in the interval the current value of the counter is displayed along with the date
and time when the counter was checked. To determine the amount the counter incremented, the delta value,
in the interval the current value must be subtracted from the previously recorded value.
For example, the following show logging onboard error-stats output shows that when the counter was checked
at 01/12/18 11:37:55 the timeout-drop counter, F16_TMM_TOLB_TIMEOUT_DROP_CNT, for port fc1/
was a value of 743. The previous time it incremented was 12/20/17 06:31:47 and it was a value of 626. This
means that since error-stats interval is 20 seconds, between at 01/12/18 11:37:35 and at 01/12/18 11:37:55
the counter incremented by 743 – 626 = 117 frames. There were 117 frames discarded at timeout-drops during
that interval.
switch# show logging onboard error-stats
----------------------------
Show Clock
----------------------------
2018-01-24 15:01:35
---------------------------------
Module: 1 error-stats
---------------------------------
displayed are not the current value of TxWait, but only deltas from the previous 20 second interval. If
TxWait incremented by the equivalent of less than 100ms there is no entry.
displayed are not the current value of RxWait, but only deltas from the previous 20 second interval. If
RxWait incremented by the equivalent of less than 100ms there is no entry.
Guidelines and Limitations for Congestion Detection, Avoidance, and Isolation
Cisco MDS 9000 Series Interfaces Configuration Guide, Release 8.x
161