Cisco Cat 3850 restart issue

[A Masch]

Hi,

I would need your help to understand the following issue:

I have 2 stacked 3850s connected to 2 Alcatel 7750 (1st 3850 has 1 link to 1st Alcatel; 2nd 3850 has 1 link to 2nd Alcatel), running bgp between them with graceful restart enabled. SSO is working.

Problem: If I restart the active 3850 it takes up to 2 minutes for the switch to be reachable again. 

During restart I see all ports go down on the 3850. After aprox 10 sec from restart, the bgp neighbourship is established, I can see graceful restart parameters as advertised / received and routes received from the Alcatel, but nothing is advertised back. After another minute the 3850 finally advertises his routes to the neighbour.

My questions are if it is normal for the ports to go down during restart and why it takes so long for the switch to advertise routes?

[Philip D'Ath]

The ports need to go down during restart, otherwise something that checks port state might start forwarding packets thinking the switch will do something with them.  So you definitely want the ports to stay down until the switch is ready to forward those packets.

How are you restarting the active 3850?  Are you sure you are not rebooting the entire stack?

[A Masch]

Hi Philip, I tried the command "reload slot <slot nr>" and also tried disconnecting the power cables from the first slot (to simulate a switch failure).

[Philip D'Ath]

What software version are you using on the 3850's?

[A Masch]

it is ios-xe 3.3.5

[Philip D'Ath]

The code release is a bit old.  Can you upgrade to the gold star release 03.06.04?

[A Masch]

The upgrade is not my easiest options.. I would prefer to keep this version and try to make it work before going for another software release.

[Philip D'Ath]

I had a lot of problems with those older code releases, such as what you are using.  A lot of issues were resolved by the time the gold star was awarded to 3.6.4.

I don't personally think you will resolve it, and I wouldn't be surprised if you run into additional issue as well.

[A Masch]

Thank you for this information. I will see if I can get them upgraded and reply with the results.