cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
2341
Views
5
Helpful
5
Replies

OutDiscards in 2960S on a 10 Giga interface

danny4569
Level 1
Level 1

 

Hi.

I have a cisco 2960S  24 x 1 giga ports with an uplink of 10 Giga connected to a IBM server.

When I am sending jumbo frames (9000 Bytes)  from up to 6 end devices (connected to 1 Giga each), there are NO packet drops.

When I connect one more end device there are a lot of OutDiscards packets on the Ten Giga ethernet interface. (under "show interface counters errors")

The total amount of the 7 end devices is less then 2 Giga bit.

I am not using QoS, all traffic is being sent with no priority, interfaces are configured as:

 

switchport mode access

switchport access vlan 2

spanning-tree portfast

 

All interfaces ang end devices are configured with auto speed and auto duplex (default)

under Vlan 2 it was configured "mtu 9000"

 

Any idea why 6 sends packets OK and 7 devices starts to discards?

Is there a way to increase the OUT BUFFER of the Tengiagbit ethernet interface? 

Regards,

Danny.

 

1 Accepted Solution

Accepted Solutions

Joseph W. Doherty
Hall of Fame
Hall of Fame

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Yea, it seems it shouldn't be possible for 7 gig ingress to have issues with 10 gig egress, but the problem is, the multi-port ingress frames/packets can arrive concurrently (i.e. and need they to be buffered).  Using Jumbo frames requires additional buffer space.

You're most likely to encounter buffer resource issues with QoS enabled and with default buffer settings.  It's less likely to happen with QoS disabled.  It may be possible to mitigate with QoS enabled with custom buffer settings.

View solution in original post

5 Replies 5

InayathUlla Sharieff
Cisco Employee
Cisco Employee

Danny,

Output drops on the port-asic always indicate you have oversubscribed the port-asic buffer space. 2960 uses 4 output queues to provide some QoS functionality in hardware at wire-speed. If you have 'mls qos' configured without any further port level config, you will find the available buffer space will be split between these 4 queues. Each queue will receive certain DSCP/CoS values and the queues will transmit in a shared round robin fashion.

Essentially::

- is 'mls qos' enabled?

- if yes, is it needed?

- if yes, can we make changes?

If you dont have 'mls qos' enabled, then you need higher linespeed as you are experiencing micro-burst that doesnt get picked up over a 30 second period. Its frustrating to find this scenario, however as mentioned, port-asic drops are simply oversubscription of buffer space over _some_ period of time. Often output drops such as yours will result from very short bursts of traffic.

2960 uses the same QoS architecture as 3560 and 3750, hence the following excellent guide is very relevant and applicable:

http://cisco.com/en/US/products/hw/switches/ps5023/products_tech_note09186a0080883f9e.shtml

 

Another excellent reference: https://supportforums.cisco.com/docs/DOC-8093

Regards

Inayath

*Plz rate if this info is helfull.

Hi Inayath

Thank you for your reply.

"mls qos" command is not enabled.

I did not need it.  There was no need until now.

It is possible to make changes!

Can you help with the relevant commands in order to change the buffer size?

Best regards,

Danny.

 

Hi Inayath

It looks that problem was solved by entering the commands:

(config)# mls qos

(config)# mls qos queue-set output 1 threshold 1 1000 1000 50 1000

(config)# mls qos queue-set output 1 threshold 2 1000 1000 50 1000

(config)# mls qos queue-set output 1 threshold 3 1000 1000 50 1000

(config)# mls qos queue-set output 1 threshold 4 1000 1000 50 1000

 

 

​Thank you for your help and excellent links. 

Best regards,

Danny.

Danny,

Glad to hear that. thanks for the rating and comments.

regards

Inayath

Joseph W. Doherty
Hall of Fame
Hall of Fame

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Yea, it seems it shouldn't be possible for 7 gig ingress to have issues with 10 gig egress, but the problem is, the multi-port ingress frames/packets can arrive concurrently (i.e. and need they to be buffered).  Using Jumbo frames requires additional buffer space.

You're most likely to encounter buffer resource issues with QoS enabled and with default buffer settings.  It's less likely to happen with QoS disabled.  It may be possible to mitigate with QoS enabled with custom buffer settings.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Innovations in Cisco Full Stack Observability - A new webinar from Cisco