You can disagree without ad hominem. Your points make little to no sense, even beyond your sneer. Pods can’t be “upgraded” individually becuase pods are epheme

You can disagree without ad hominem. Your points make little to no sense, even beyond your sneer. Pods can’t be “upgraded” individually becuase pods are ephemeral. Downgrades work fine - replace newer binaries with older binaries. Upgrades work great too - replace older binaries with newer binaries. Single node is the correct context for upgrades because a node is the failure domain in kubernetes (and in every single other Linux based orchestration system). 1.28 even includes an automated way of doing this in a highly available way per host. CSIs and CNIs are binaries, which can be upgraded and downgraded like with any other system. Yes, if you upgrade or downgrade major versions there may be breaking changes. I’ve been a sysadmin and a devops engineer and a cloud platform engineer for approaching 20 years. I’ve flagged your reply as I don’t think it’s appropriate, but I wanted to reply to you anyways because I’m sure beyond your anger and rudeness and lack of empathy, you’re intelligent enough to do a bit of reading and double check that you’re not absolutely wrong on all counts. At some point you’ll be surrounded only by what you think are idiots, and they’ll exclude you. At that point you might investigate if just possibly you’re not by far the smartest person on earth. It’s possible other people know what they’re doing too.