Monitoring LC resources usage

[ThomasD86]

Hi,
in our network we're using the A9k-48x10-1G-SE LC on our ASR 9k deployment (running eXR 7.3.2), these boards have huge (in the thousands) l2transports interfaces configured on them and, most of these, have a service-policy associated.

Is it possible to monitor what is the current status of the resources the NPUs on the card have available to handle qos policies? I am assuming that a hierarchical policy that has a grandparent-parent-child would consume more resource than just a standard one, is there a CLI to see that?
 
I've found the "show qoshal resource summary location X/X" CLI which seems it should provide the info but, the output is not easy to interpret and I couldn't find any documentation available to make sense of it.

What would be the best way to monitor the resource availability on a tomahawk LC?

Thanks for the help

[anpetit]

Hello @ThomasD86 , 
Checking your post, "show qoshal resource <np x> summary location X/X" is the CLI you are looking for as you mentioned. 
Most important information from this CLI is:
- The QOS-EA, which display the resources used on the LC regarding QoS.
- Policers which display the number of policier configured.
- Then I assumed you are also observing (L1,L2.. L4 entities). 

You can also use "show controllers np counters <np><location>" and check for "queue tail" and "wred" counters. 

Hope it helps. 
Kind Regards,
Antoine

[anpetit]

Hello @ThomasD86 , 

Hope you are doing well. 
Let me know if you need more help regarding this otherwise can we close this thread out ?
Have a nice day. 
Kind Regards,

Antoine

[ThomasD86]

Hi Antoine,

thanks for your reply. I am well thank you, hope th same goes for you.
I apologize for my late answer but I was trying to figure out the output, I took the attached output from one of the busiest router in the networks (redacted to include only qos_ea) and while I can see how many policers we do have configured, what is not clear is how can I understand from that output how much more resource for qos-policing NPUs have still available for policers configuration. I can see that on NP0 I have configured 2.1k policers but how to determine how many more we can configure before the NPU is out of resources?

Looking at the NP counters, there was no item matching "queue tail/wred"

Thanks for your help so far!

Regards

---

---

[anpetit]

Hello @ThomasD86 , 

Thanks for your reply.

Just sharing the thread below, you might be able to get a better look for your info. 

https://community.cisco.com/t5/xr-os-and-platforms/asr9k-bng-bundle-ether-interface-policers-duplication-for-pqos/td-p/3675317

Let me know if it helps. 
Have a nice day.

Kind Regards,

Antoine