New Windows Servers are like buses: None for ages, then two at once! One of the previews is for the Windows Server vNext Long-Term Servicing Channel, aka Windows Server 2019. The other previews the Windows Server Semi-Annual Channel, the version that gets a release every six months but is only supported for 18 months. Both previews receive “in-place OS upgrade” from Windows Server 2012 R2, …
i was coming to say the same thing. Forced in place upgrade with exchange installed? errrrrr no. Our exchange server is still running 2010 on 2008r2. it was scheduled for side by side migration to 2016 on 2016 this summer. Ill have to ensure it IS scheduled for this summer if we are being forced onto this 18 month inplace malarky.....
wait until the equivalent of GWX "happens" - your exchange mailboxes suddenly 'migrated' and you're like W.T.F. just happened? [especially when it breaks and/or you get a 2D FLATSO TheMetro config screen]
just sayin'.
FYI - I migrated several-year-old Cyrus IMAP on a FreeBSD server late last year with no hitches at all. Had to replace the hard drive, that's why. it's all running happy on ZFS with compression and replication enabled, and no data losses, EVAR.
just sayin'. again.
(so, WHY do you need Exchange?)
Speaking about extinct SKU's, can we get some love for Foundation Server? (Almost) everything Standard server could do, none of the extra stuff that can cause problems in Essentials/SBS, pre-licensed for 15 users, and in place upgradable to Standard once you reach that point.
It was the perfect "My First Server" for SMB Customers.
Like most, I always prefer to do a completely fresh install over any kind of in-place upgrade.
But sometimes the customer gets the say and don't want to go down that route, despite all the best arguments otherwise. Usually for silly reasons like legacy software where they've misplaced the install media etc.
But to then say that it could trash domain controllers...is...well it's typical MS lately, to be fair.
Of course, things like this are exactly why we have Technet so we can try it in our lab first...oh...yeah....
i have performed a few inplace upgrades for features. One was a simple file server and the customer wanted deduplication. Ive also inplaced a 2008r2 to 2012 that had sql server 2008 to 2014 followed by a 2008r2 to 2012 (!!) as the customer had no alternative hardware but was told a vital web app was not supporting 2008sql or 2008r2 with a short deadline. So we patched, cold isolated the server, backup up and went for it. It worked first time and that was a few years ago.
I would never inplace a domain controller (id promote another, demote the inplace first if i HAD to). And exchange? hell no. Domain enterprise certificate server? Yikes! Cluster host? now you are the realms of silly talk.
The only time you would ever do an in place upgrade of an NT based OS (outside of a test environment) is when you upgraded from an NT 4 domain to AD on Windows 2000. You had no choice, you had to upgrade the PDC to Windows 2000. Of course you then immediately built a new DC, transferred the FSMO roles and nuked the upgraded server.
Not so much call to do this nowadays.
"IIRC you have to do the first one in each domain. You can then join other new version hosts and promote them, move roles, and eventually demote and remove the in place upgraded DC."
No you don't. Build the new one, join to domain, promote to DC, rinse, repeat, retire old ones, raise AD functional level.
In-place upgrades... fine so long as you use VM's:
VMs:
Checkpoint.
Replicate.
Backup.
Remove the replica.
In-place upgrade the replica, put it on a test VLAN as you do-so.
Make sure it works.
If so, either downtime and then replace production with upgraded replica, or risk an in-place upgrade of production.
Hypervisors:
Replicate all VMs to somewhere else.
In-place upgrade.
Test.
Replicate them back.
It's 2018. Use VMs. They save your life in these situations. (I would suggest that you shouldn't need to backup hypervisors as they shouldn't contain anything over than a basic Server install, and copies of the VMs. Maybe a backup agent. That's it. Should you lose a hypervisor, it's shouldn't ever be a worry to just wipe it out and start again, it'll probably be quicker than a restore to be honest).
But nobody should be in-place upgrading any production server whatsoever without VM's.
P.S. if you're not already VM'd.... do it. You get (is it 4?) free VMs for every version of server. Literally, virtualise the server, put it back on itself running as a hypervisor. Done.
Take any mention of features with a bucket load of salt.
Say for example the Semi Annual channel where there were due to be enhancements to S2D but it was disabled in the 1803 release to make it 'container focused'.
No mention of this anywhere except the final release notes, all the tools think it works, but the functionality has been hacked out. Mildly annoying especially after all the publicity and previews.
I really don't think they properly understand that LTSB and semiannual should be different release tempos and support lifespans and *NOT* different products with one getting new features but crippled for most purposes, and the other more or less abandoned for feature updates for life.
All smells very political and artificial which is just what I expect from the current MS leadership.
It's a model that works quite well. Any business with a sane IT plan will be using LTS.
Hopefully software developers now get to fix things properly with a couple of years grace before any changes are rolled up into the next LTS.
But how did the 'In-place DC upgrade' ever get approved as a 'feature', this has to be a prime indicator of the complete loss of real world understanding in the MS Ivory towers.
yuk yuk. Nurse, nurse, etc.
See these scars? Previous employer with most of the prod infrastructure running on Red Hat 8 earlier this decade, because dev claimed it would take the entire team working flat out for two years to port it to a supported flavour.
No, no, you misunderstand. RED HAT 8. Not RHEL... it had been unsupported for best part of a decade by then, IIRC. Saved an awful lot of time that would have been wasted writing test cases for regression testing, I'm sure.
(I'm posting this from a Linux laptop that I'm supposed to be using to review the WS2016 Security Guide...)
This post has been deleted by its author
The above phrase caught my eye.
So let me get this straight: having found nothing screwed up, are MS giving their software carte blanche to actively screw things up? Or is there a "No Thanks" button one can press?
Will you start off with a Network problem and end up with the same Network problem together with a non-functioning USB printer, for example?
(Current Technology Strategy: Whenever a user reports an issue where a "diagnose problem" box appears, my first question nowadays is "Did you Ok that box?" If they admit to it (not a problem - so long as I know), then I start by looking at what that diagnostic changed, rather than assuming no settings had been changed.)
Anyone else heard the recent news that Microsoft is discontinuing most of the fonts included with any of their software and are going to a paid model? Fonts will no longer be developed as the old TrueType type and will come from the Microsoft Store, licensed separately, with separate fees for anything you fancy.
It's true. The current leadership is so subscription-fee driven that they won't even include a few things formerly provided and expected. I haven't paid for fonts since the early '90s and I don't think I'm going to start doing that again.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
