3850 Interface Transmit Drops

[groupccologin]

Hope you can help me

We have moved our Distribution traffic from a Pair of stacked WS-C3750G-48TS switches to a Pair of Stacked WS-C3850-24T late last year. All switch interface configs were moved like for like from the old stack to the new stack and brand new Cat6 Cabling  . All appears to be working fine with zero latency or performance issues .

-3850-24T software is Version 16.9.4 and seems stable

Since moving the hosts we have noticed the interface counters have been showing high transmit drops in Solarwinds and on switch even when resetting counters . 

Has anyone seen this behavior before ?

As i said performance and port config all seems fine with no problems being reported, Software bug?

Any help appreciated

Damian

[Reza Sharifi]

As long as you don't see any transmission drops on the actual interfaces of the switches, you should be fine as this is most likely a bug in the version of the software you are running. You could also check with SW to make sure you have the latest, bug free NPM software.

HTH

[groupccologin]

Thanks for the Reply 

We arent seeing any transmissions drops at all interface level and general performance seems excellent on the stack .

Alongside this we are looking at Solarwinds NPM builds to see if there any bug fixes on the version we are on ?

Cisco software version information on this seems lacking in the software release notes for the next version up on our release and  Im hesitant to just start trying new software versions to mitigate this at a switch level.

[Reza Sharifi]

Cisco software version information on this seems lacking in the software release notes for the next version up on our release and  Im hesitant to just start trying new software versions to mitigate this at a switch level.

I completely agree. Since everything is working fine with the switches, there is no need to start making any changes by upgrading or downgrading software versions. This could simply be a cosmetic bug.

HTH 

[Kevin_W]

Hi groupcclogin,

besides the packet loss itself it would be nice to know how much percentage the packet loss is (("total output drops" / 100) * "packets output"). Maybe it is e.g. under 0,1% and can be unattended if there are no impacts reported.

[Leo Laohoo]

Post the complete output to the command "sh run | i qos queue-softmax-multiplier 1200".

[groupccologin]

Thanks for the feedback

Running the following :

sh run | i qos queue-softmax-multiplier 1200

Produces no output just puts me back at the cli prompt

Damian

[Leo Laohoo]

Put the command "qos queue-softmax-multiplier 1200" into your switch.

Next, observe if Total Output Drops counters have stopped incrementing or incrementing very slowly.

[groupccologin]

Can i ask what this command does before i enter it ?

Does it need to be put in at the config elevated level?

Damian

[Leo Laohoo]

---

@groupccologin wrote:

Can i ask what this command does before i enter it ?

Does it need to be put in at the config elevated level?

---

I don't want to send people into a coma shock by "explaining" it.  Read it:  Tweaking Buffer Allocation

When using 3650/3850 (maybe IOS-XE), enable this feature and reboot the switch/stack.  

[groupccologin]

Thanks for the link  I’ll have a read make sure it’s bedtime reading 😂. Very helpful thread appreciate the feedback from everyone 

[groupccologin]

I’ve applied the command Suggested all seems well still with traffic across the stack flowing well and low latency 

The port counters are still incrementing though so unsure of the reason  . Going to raise a TAC case now hopefully they will shed some light. Thanks for the pointers hopefully I’ll get to the bottom of it .