6

In the spirit of asking a Good subjective question I bring this forth again, in hopes that ServerFault will continue to be a preferred "go to" when searching on this topic.

OLD REFERENCE

Back in 2011 this question was asked: Reasons Why In Place Upgrades Are Bad

THE QUESTION

Is doing an in-place upgrade of Windows Server x64 (2008 R2, 2012, 2012 R2) to a newer supported Windows Server version a good practice these days or no?

THE ANSWERS

Remember from the link above: Great subjective questions tend to have long, not short, answers. The best subjective questions inspire your peers to share their actual experiences, not just post a mindless one-liner or cartoon in hopes of being rewarded with upvotes for being merely "first."

Community
  • 1
TheCleaner
  • 32,627
  • 26
  • 132
  • 191

3 Answers3

9

My opinion is, when supported, an in-place upgrade of a Windows Server OS to a newer version is the preferred method for upgrading a server (especially a VM) these days. Not only is it much quicker, and allows for greater automation, but it also no longer holds as much risk vs. reward that older OS in-place upgrades represented due to Microsoft's support of a better lifecycle transition model.

Microsoft has done a lot to make sure that in-place upgrades are easier than before and more seamless, without the problems that plagued older OS upgrades. This seems to be the recommended course of action if you want to keep everything as is: Windows Server Installation and Upgrade:

"If you want to keep the same hardware and all the server roles you have set up without flattening the server, upgrading is the way to go"

Even MS bloggers are believing the hype: In-Place upgrade of 2008 R2 to 2012 R2

In addition, VMs allow for snapshots, P2V, cloning, and rollback to ease in this transition. (Reference: "P2V it into a test VM and, uh, test it."

We should be "Treating servers like cattle not pets" - Randy Bias. The old days of caring and feeding the servers you are responsible for isn't done as intimately as it once was. Not every server is a snowflake, and public cloud hosting is a prime example of this practice.

For instance:

I've personally done multiple in-place upgrades from 2008 R2 to 2012 R2 over the past year, with about 400 more planned for the next year and a half. All have gone just fine, with only minor issues afterwards on 2 servers that didn't require rollbacks. As long as the existing server is running fine, you should feel confident in moving this direction.

Some important things to consider when making this claim that in-place upgrades are the preferred way to go:

  • Is the upgrade supported? - For example, switching language versions, or build types isn't Upgrade Options to Windows 2012 R2
  • Are the applications running on the server fully compatible with the new OS version in their current state?
  • Is there a way to rollback easily to the previous state before the in-place upgrade (VM Snapshot, backup, clone, etc.)?
  • Does the hardware (if applicable) support the new OS?

  • Is the existing server running well currently without issues?

    If the answer is YES to ALL of the above questions, then an in-place upgrade is the preferred route.

Community
  • 1
TheCleaner
  • 32,627
  • 26
  • 132
  • 191
  • `Even MS bloggers are believing the hype` wouldn't they be the very ones to push the hype, regardless of merit, in the first place? – SnakeDoc Jun 20 '16 at 16:47
  • And frankly, it's not whether or not an in-place upgrade works, it's about an in-place upgrade leaving garbage everywhere and generally not doing a good job at a clean upgrade. Just look at the recent upgrade-to-windows10 snafu that people are still running into. Half uninstalled old versions of things, scheduled tasks that don't execute, forgotten file associations, orphaned files strewn everywhere, etc. If you have an important system, it's still best to fresh install rather than upgrade. – SnakeDoc Jun 20 '16 at 16:51
  • 1
    @SnakeDoc - 1) we are talking about server OS' not client side...which should be better controlled. 2) servers should be kept in good working shape (my final bullet point), if they aren't then doing an upgrade won't suddenly cure the issues it already is having, and may point them out even more. 3) Try migrating hundreds of servers these days, you'll be doing it for years. The upgrade model has matured enough to be the go to now. I'm completely fine with migrations when warranted, but you'll have to prove out your "it's still best" point with citations and reasoning not just conjecture. – TheCleaner Jun 20 '16 at 17:27
  • Onus of proof is squarely on you here. You're making the claim against what most people consider is standard practice. The reasons an in-place upgrade are still not optimal are very well laid out in the old post you linked to. None of the reasons for why an upgrade is bad has changed since that answer's writing. You have debunked nothing except state it saves time. A server OS in Microsoft World is essentially the consumer system with additional services. So, no, I don't accept Microsoft have perfected their upgrade system only on their server products, but left all others to rot. – SnakeDoc Jun 20 '16 at 17:31
  • I gave you "proof" with multiple citations of my opinion in my answer. . You have your opinion, we can leave it at that. My citation with quote is what MS themselves state. Whether you agree with it or not doesn't change what their current recommendation is. You are welcome to post an answer as an option. Most of it is a gray area where the local choice is the best choice, and I could have easily argued the other direction, but I think the times of keeping a server running on the same OS for 10 years is gone and in-place upgrades are more viable than migrations these days. – TheCleaner Jun 20 '16 at 17:39
  • None of the reasons stated here: http://serverfault.com/a/224847/164245 have been addressed. And they are likely to never be addressed. Until that day, wipe and reinstall will always prevail, especially in Microsoft World where installed packages have a tendency to cling all over the system instead of living neatly inside a single directory. What you have laid out is a case for lazy system administration, which is fine by itself, but unfortunately doesn't apply here. Just because a vendor recommends something, or provides a tool to do something, doesn't mean that's the best course of action. – SnakeDoc Jun 20 '16 at 17:44
  • In-Place upgrades as the recommended practice by the OS OEM themselves isn't being lazy. Going down the old school path that we've all been down of migrating is still a viable option and one you can choose, but I contend that most have better things to work on than keeping an OS current. If you have the time and resources to migrate thousands of servers like Phil did, then yes it is the best option. Most IT shops don't have that luxury or spend. But I digress, I'm not going back and forth with you over something that you can simply downvote and call it a day on. – TheCleaner Jun 20 '16 at 17:58
  • Bottom line is you're trying to justify something (if not just to yourself) that most others wouldn't do. The reason why they wouldn't do it are clearly laid out, and you even linked to them. Not a single thing has changed to somehow make in-place upgrades preferred or normal practice. Microsoft has said from day 1 that in-place upgrades are recommended... but the industry found out otherwise in practice. Do with your systems what you will... to each his own, etc. – SnakeDoc Jun 20 '16 at 18:07
  • The real bottom line is that you want to argue this like it's religion. The whole point of a subjective question and answer is justification. There isn't a clear mathematically algorithm here. Lots of things have changed in the last 5 years to make it the preferred method. The industry is changing rapidly and "otherwise in practice" is only good until the paradigm changes, which I believe it has with OS upgrades. I'm not going to tell my admins to migrate 3,000 servers when 95% of them will upgrade without issue just because that's how we used to do it in the 3.11/NT/2000/2003 days. – TheCleaner Jun 20 '16 at 18:26
  • The real thing that has changed since the 2003 days is automatic provisioning and rollout. This alleviates the issue you are up against, and makes your systems disposable (you can nuke and pave any system at any time for any reason, and it will come back up exactly as it should). That's what big shops aught to be doing... not in-place upgrades. Justify it as you will, but 6 months from now, after you've re-installed half of those systems due to random issues... please post an update for the rest of us. – SnakeDoc Jun 20 '16 at 18:33
  • I have in-place updated various systems without issues, I do prefer in-place upgrades as well nowadays. If issues arrive, the box can still be nuked and rebuilt – MichelZ Aug 23 '16 at 17:08
0

Just for on-going reference, and to add detail straight from the horse's mouth from a more recent article, Microsoft does not support Azure upgrades. Cattle are killed, not upgraded.

Andrew Schulman
  • 8,811
  • 21
  • 32
  • 47
jshessen
  • 1
-3

I would like to point you to Microsoft's Office 365 fundamentals course. They explain how Microsoft stamp out a bunch of server with automated build processes. They don't upgrade or even patch them, they just stamp out another bunch of servers using automated deployment tools. Automation is king and I need to get better at it.

Daflibble
  • 1