Junos OS

 View Only
last person joined: 8 hours ago 

Ask questions and share experiences about Junos OS.
  • 1.  Junos Upgrade QFX5110 and EX3400 in VC

    Posted 09-23-2023 20:00

    Hi all,

    I need to upgrade my switches soon with minimum outage downtime. I need your advice on this. See below for the VC setup. I would like to upgrade them to a recommended version. So which method should I apply it to do the upgrade. I understand in various discussions in the community that you should not use NSSU, because it can cause many issues during the upgrade procedure and it can cause a longer down and upgrade time with low success rate. Another method is to upload the junos image on master node and have it upgrade the rest of the switches and then reboot all switches member "request system reboot all-members". This will create a full downtime with less time than NSSU, so is probably better scheduling a maintenance window. Possibly the last method is to perform a manually upgrade by breaking the VC to upgrade each switch individually. So first upgrading the linecard nodes and after the backup and master node. So can anyone suggest the best method (best practice) to upgrade the QFX5110 and EX3400 in VC.

    VC setup:

    4 member of QFX5110 in VC (20.4R3.8 to 22.2R3-S2)

    2 member of QFX5110 in VC (20.4R3.8 to 22.2R3-S2)

    2 member of EX3400 in VC (20.4R3.8 to 21.4R3-S4)


  • 2.  RE: Junos Upgrade QFX5110 and EX3400 in VC

    Posted 09-28-2023 13:07

    I would recommend upgrading the software on all VC members at the same time in the MW.  Breaking the VC just to upgrade may not be a good idea. Involves lot of overhead in removing the switch and adding it back.

    Apurva Dipak Modh

  • 3.  RE: Junos Upgrade QFX5110 and EX3400 in VC

    Posted 10-04-2023 12:56

    Thank u guys for the advice. i ll keep that on mind.  I see most people also recommend the same solution as in upgrading all VC member at the same time in the maintenance window.

  • 4.  RE: Junos Upgrade QFX5110 and EX3400 in VC

    Posted 10-02-2023 18:37

    The procedure all depends on what redundancy your connected devices have. If they are mainly hypervisors and other devices that are multihomed, this can be done quite nicely. The more single-attached devices you have, the longer the downtime.

    I guess you have already read this guide:


    It is easier to handle 2 member VCs in this scenario. Perhaps splitting the 4 member VC into two VCs would be a good idea.

    Another idea is to go for eVPN on the QFXes. This may require a license upgrade, but gives you superior flexibility and any (well, most) upgrade issues will become a memory only: