Upgrading Virtual Chassis of EX8200 with minimal downtime
I have to upgrade a Virtual Chassis composed of 2 XRE-200 and 2 EX8216 from version 11.3 to version 12.3 with minimal downtime.
ISSU is out of the question because version 11.3 does not support it, should be in 11.4 for a proper ISSU (according to JTAC).
I am considering splitting the Virtual Chassis in 2 groups:
Member 8 (Primary XRE) and Member 0 (EX8216)
Member 9 (Secondary XRE) and Member 1 (EX8216)
This split would consist in the following:
Disable all interfaces existing in Member 1 (I have full redundancy on the links)
Take down all VC connections between both groups.
Maintain only a single VC connection between XRE and corresponding EX8216
I am planning to upgrade the first group that had the standby XRE (Member 9 and Member 1) to the new version.
After rebooting the group and verify its sanity, I'll disable all interfaces in the the Master group (Member 8 and Member 0) and activate traffic on the new upgraded group (Member 9 and Member 1). This is where I'll have downtime
Then upgrade the remaining group (Member 8 and Member 0), reboot, shutdown, connect all vc cables and then power on again to integrate correctly in the VC.
I have a doubt in the process that I cannot test because I don't have a lab with such boxes.
On the upgrade of each group, there will be only one connection to a RE of each EX8216 to the corresponding XRE200.
On the Secondary XRE, this will be connected to the Backup RE of the EX8216.
Do you know if when I upgrade a group of XRE200 and EX8216, both RE of the EX8216 will be upgraded despite one of them not being connected through Ethernet?
I am most certain that it will be fully upgraded but I'd like your input.
I have contacted JTAC but I am still struggling with them fro them to understand what I want.
Re: Upgrading Virtual Chassis of EX8200 with minimal downtime
Thanks for the clarification.
I still don't know if we will go down that road since it will be customer decision. JTAC does not recommend neither supports this kind of upgrade.
They suggested they had problems with duplicate IP addresses and problems with LAGs after activating the other RE but I suspect that in their method, the VC cables were not disconnected to isolate physically both halves of the VC.
I strongly believe that my method should work but unfortunately I don't have the equipment to test and applying it in a critical production environment might not be the ideal subject to confirm it. 🙂