Typical leaf upgrade impact

[Dan Peronto]

This will be my 1st time upgrading the fabric in production.  We do have some single connected devices that will be impacted.  Does anyone know the typical amount of time the leaf will not foward packets during an upgrade?  Going from 1.3(2h) to 2.0(2m) if that matters.

[Manish Gogna]

Hi Dan,

As per the following

http://www.cisco.com/c/en/us/td/docs/switches/datacenter/aci/apic/sw/2-x/managing_ACI_fabric_upgrades_and_downgrades/b_Managing_ACI_Fabric_Upgrades_and_Downgrades/b_Managing_ACI_Fabric_Upgrades_and_Downgrades_chapter_01.html

While a single maintenance group will upgrade all leaf and spine switches at the same time, Cisco recommends that you divide your leaf and spine switches into multiple (two or more) maintenance groups to prevent the entire fabric from going down during a software upgrade. Dividing up the leaf and spine switches into two or more maintenance groups, composed of roughly equivalent groups of leaf and spine switches, allows continued operation of the fabric during software upgrades by upgrading half (or less) of the fabric nodes at one time.

The switch upgrade takes up to 12 minutes for each group. The switches will reboot when they upgrade, connectivity drops, and the controllers in the cluster will not communicate for some time with the switches in the group. Once the switches rejoin the cluster after rebooting, you will see all the switches listed under the controller node. If there are any VPC configurations in the cluster, the upgrade process will upgrade only one switch at a time out of the two switches in a VPC domain.

HTH

Manish

[Dan Peronto]

Is packet flow only interrupted during the reboot portion of the upgrade or the entire 12 minutes?