I have a pair of C9300-24T switches running as a stack and where all connections use symmetric LACP etherchannels.

Is it possible to do a software upgrade without downtime? (Meaning that one side of each etherchannel is up at all times).

Current version is 16.06.03. Following the documentation when upgrading to that version (install activate on the master switch) caused both switches to reload at the same time.

1 Answer 1


If it is a full IOS-XE upgrade, then no, you will need to restart the stack to run the new code. Remember that all the switches in the stack must be running the same code version.

If it is a SMU (Software Maintenance Upgrade), then it may be possible to upgrade a module without restarting the stack, but you will probably have some downtime, depending on the specific upgrade. The 9300 series was designed so that separate modules of the code can be individually upgraded, but as far as I know, Cisco has not yet released any such upgrades.

In any case, you should always anticipate needing to have some downtime when performing a software upgrade on a switch or switch stack (including switches that support ISSU), even if it is only a few seconds. You should only do such work during a network maintenance window.

Cisco has many documents regarding upgrading software on the 9300 switch series if you simply search.

  • Thanks, I was afraid that was the case. I have found some 9300-specific Cisco documents that mention a procedure for ISSU and some that say it's not supported, so I thought I'd ask. Even one or two ten second interruptions would be acceptable if there's an emergency security fix, but we'll just have minimize the attack surface instead so we can justify waiting for a nightly window.
    – LHMathies
    Jul 9, 2018 at 12:29
  • 1
    ISSU is not supported for SMU. We have used ISSU with dual-supervisor 4507 switches, but you still lose connectivity for a brief period (sometimes only a second or two), although it is much less time than rebooting the entire switch.
    – Ron Maupin
    Jul 9, 2018 at 12:34
  • 1
    It feels a bit strange to have spent a lot of money for a redundant stack, especially one where the supervisor in one unit actually runs the hardware of both (as I understand it) -- and then find it's not able to upgrade the idle supervisor and swap over to upgrade the other. Luckily it works fine on ASAs.
    – LHMathies
    Jul 9, 2018 at 12:53
  • I think once Cisco gets all the SMUs going, then things will be better. I saw a list of what requires a restart, and what does not. I suppose they will explain that for each SMU released. Also, the 10 year support is something we have wanted. It is tough that by the time we get a model through the lab and want to start deploying it, the device is only a couple of years from EoS. (We have many tens of thousands of switches.)
    – Ron Maupin
    Jul 9, 2018 at 12:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.