port tracking

[Mahmoud]

Hello,

lately we have a flapping uplinks from one leaf (201) to spines (101&102). all of our endpoints are connected with vPC architecture to the (201,202). all endoints that connected to the affected leaf suffered from traffic dropping inspite it connected to other leaf that working normally

port tracking is disabled so I ended up with enabling this feature

do you have any recommendation for this?

Is there any impact to enable this feature?

[Sergiu.Daniluk]

Hi Mahmoud,

You need to be very careful when using port tracking feature. Reason is that if the policy is triggered, ALL leaf switch access ports are shut down. So if you have any single-home connected servers/devices on the affected switch, they will be impacted.

If you set the threshold too high, you might create more problems then you solve. For example if you set the threshold to 1, but you have only 2 spine links, then if a spine switch reloads, your ENTIRE fabric will go down.

Cheers,

Sergiu

[Mahmoud]

Hi Sergiu

thank u so much for your valuable feedback.

I am planning to use the threshold to be zero so when leaf loose all the connectivity with spines it will disable all downlink interfaces with connected endpoints.

correct me if I am wrong.

also do you know is there any service interruption will happen when we apply this configuration?  as no documents from cisco talking about this feature in detail

[Sergiu.Daniluk]

Hello,

Yes, by enabling the feature with the threshold 0, then leaf switches will bring down access ports when all fabric ports are down.

I do not believe there is any impact when enabling the feature. However, I would suggest to enable fabric wide settings only outside of working hours.

Also, it seems that it is recommended to have the feature enabled:

> Except for very specific server deployments, servers should be dual-homed, and port tracking should always be enabled.

Ref: https://www.cisco.com/c/en/us/solutions/collateral/data-center-virtualization/application-centric-infrastructure/white-paper-c11-737909.html#_Toc6452913

Cheers,

Sergiu