New Server Migration Adventure



  • Several years ago on ML, I documented my adventure of a server upgrade and Windows Server migration. Primarily, the challenge was upgrading the AD server from Small Business Server 2003 to Server 2012 R2. That was a challenge but ML was instrumental in helping me find the best documentation and methods to do so. I'll never forget the night that I pressed enter on the AD migration script but it went 100% perfect.

    Since then, via Hyper-V, we run our AD, File Server and Backup Server (Veeam and BackupAssist) on that one physical server. I also re-purposed and purchased licenses for the old server and it also runs 3 virtual servers. All 6 virtuals are 2012 R2.

    So today, our new Dell VRTX arrived and it will replace both of those physical servers. It is a two node system based upon PowerEdge M630 with plenty of storage and a ton of drive bays left for future use.

    The purchase comes with Remote Consulting Service so I'll use that as much as I can, especially in configuring the nodes.

    I also believe that migrating the VM's won't be so bad though I am looking for best practices on that.

    We purchased the appropriate licenses for Server 2016 with Software Assurance. As mentioned, everything is currently on 2012 R2.

    The easiest plan would be to migrate the VM's over as they are and upgrade them to 2016 later (we have the license to do so.) But I'm thinking it might be best to go ahead and knock that out now...

    So looking for some opinions, suggestion and best practices...thanks...



  • This isn't a best practice, but more a personal preference. But when doing work like this, I try to avoid doing a V2V at all. Anytime that I see a V2V making sense, I consider it a failing (which happens, but it's important to identify these items.) In theory you should never "want" to migrate workloads as they are running. Sure for HA you do, but HA is a different animal even if it is a related physical mechanism.

    Even if I was just moving from 2012 R2 to another 2012 R2, I'd prefer to do a fresh build, and migrate only the data, not the running system, from one place to the other.

    Going from 2012 R2 to 2016, I would want to do this even moreso. I would want completely fresh, clean 2016 VMs and move workloads to them one by one in a pristine environment. Best to do so using a scripted or state based build process that is repeatable (and provably so by being repeated) so that you can do it anytime, over and over, whenever you need.

    It's an opportunity to make the setup cleaner and more reliable.



  • So an example, let's say you have a simple Windows Web Server... Just 2012 R2 running IIS for a .NET application or something like that.

    I'd want to create a clean 2016 build and have a script that sets up absolutely every aspect of the system without any human intervention (other than running the script, obviously) and either builds the entire system 100% or builds it to a point that all you have to do is restore the application files.



  • What kind of workloads will you be upgrading?



  • @scottalanmiller

    And I agree with that. So if I head in that direction, for sure looking for tips/tricks to make it go smoothly...

    @reid-cooper said in New Server Migration Adventure:

    What kind of workloads will you be upgrading?

    VM1: Active Directory / DNS / DHCP
    VM2: File Server / VPN
    VM3: Backup Server (just runs backup software and copies to NAS...runs Veeam / Backup Assist)
    VM4: SQL Server 2014
    VM5: Applications Server (locally hostes applications and some metered billing programs for our copiers)
    VM6: Docuware

    I'm thinking of a 1 by 1 migration to 2016, starting at VM6 and moving up and then retire the old server when VM1 is complete.



  • @garak0410 said in New Server Migration Adventure:

    VM1: Active Directory / DNS / DHCP

    Easy. Make a 2016 AD DC, add it to the pool, move roles, shut the old one(s) down.



  • @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    VM1: Active Directory / DNS / DHCP

    Easy. Make a 2016 AD DC, add it to the pool, move roles, shut the old one(s) down.

    Excellent...note - my file server also has an alias so it will help. There are a few things in our network (copiers and Docuware) that would only work with the primary DNS name, so will need to be aware of that.



  • @garak0410 said in New Server Migration Adventure:

    VM2: File Server / VPN

    Remove VPN from that. That doesn't belong there.

    File servers are pretty easy. Normally the shares are trivially easy. Script that out and just do a file transfer.



  • @garak0410 said in New Server Migration Adventure:

    VM4: SQL Server 2014

    Script a fresh install. Then just restore a backup of the old database.



  • @garak0410 said in New Server Migration Adventure:

    VM3: Backup Server (just runs backup software and copies to NAS...runs Veeam / Backup Assist)

    Fresh install might be all that is needed here.



  • @scottalanmiller said in New Server Migration Adventure:

    So an example, let's say you have a simple Windows Web Server... Just 2012 R2 running IIS for a .NET application or something like that.

    I'd want to create a clean 2016 build and have a script that sets up absolutely every aspect of the system without any human intervention (other than running the script, obviously) and either builds the entire system 100% or builds it to a point that all you have to do is restore the application files.

    I know when adding roles, you eventually will have an opportunity to save an xml config to use for redeployment.



  • Is your file server set up with two virtual disks? one for OS and the other for data only? Because will be one easier move.



  • Is there a possibility of doing a live migration or configuring Hyper-V Replication to the VRTX setup? Then you can upgrade to 2016 at your leisure.

    That being said, if your other plans include any down time I would go ahead and spin everything up a 2016 as part of that.



  • @garak0410 Assuming everything is staying on the same hypervisor platform, I would simply move your existing workloads and shutdown the old hardware.

    Then you can migrate the workloads as a separate project.

    Mixing projects is generally a bad idea IMO.



  • Thanks for all of the tips thus far. I may have to rush this migration now...

    I came into the office today with the oldest of the two physical servers sounding like jet airplanes taking off..before, it just sounded like a sick car. It could be days our hours before it may give it up. Not sure why the fans are running so high today but add that to some bad memory, failed drives (but still working RAID set)...it simply can't wait too much longer. I asked for this upgrade 6 months ago. But I digress.

    I'd still like to take time and do the migration to 2016 rather than just moving the VM's...luckily the weekend is here (though I have to coach a few ball games this weekend.) But for sure don't want to rush this.



  • @garak0410 said in New Server Migration Adventure:

    Thanks for all of the tips thus far. I may have to rush this migration now...

    I came into the office today with the oldest of the two physical servers sounding like jet airplanes taking off..before, it just sounded like a sick car. It could be days our hours before it may give it up. Not sure why the fans are running so high today but add that to some bad memory, failed drives (but still working RAID set)...it simply can't wait too much longer. I asked for this upgrade 6 months ago. But I digress.

    I'd still like to take time and do the migration to 2016 rather than just moving the VM's...luckily the weekend is here (though I have to coach a few ball games this weekend.) But for sure don't want to rush this.

    Moving from two physical machines down to one, seems like an odd move. When this one starts to die, a migration off of it will be even more dramatic.



  • @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    Thanks for all of the tips thus far. I may have to rush this migration now...

    I came into the office today with the oldest of the two physical servers sounding like jet airplanes taking off..before, it just sounded like a sick car. It could be days our hours before it may give it up. Not sure why the fans are running so high today but add that to some bad memory, failed drives (but still working RAID set)...it simply can't wait too much longer. I asked for this upgrade 6 months ago. But I digress.

    I'd still like to take time and do the migration to 2016 rather than just moving the VM's...luckily the weekend is here (though I have to coach a few ball games this weekend.) But for sure don't want to rush this.

    Moving from two physical machines down to one, seems like an odd move. When this one starts to die, a migration off of it will be even more dramatic.

    Moving to a Dell VRTX with 2 nodes...not that makes it easier to migrate off of that one day but very glad to have it.



  • @garak0410 said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    Thanks for all of the tips thus far. I may have to rush this migration now...

    I came into the office today with the oldest of the two physical servers sounding like jet airplanes taking off..before, it just sounded like a sick car. It could be days our hours before it may give it up. Not sure why the fans are running so high today but add that to some bad memory, failed drives (but still working RAID set)...it simply can't wait too much longer. I asked for this upgrade 6 months ago. But I digress.

    I'd still like to take time and do the migration to 2016 rather than just moving the VM's...luckily the weekend is here (though I have to coach a few ball games this weekend.) But for sure don't want to rush this.

    Moving from two physical machines down to one, seems like an odd move. When this one starts to die, a migration off of it will be even more dramatic.

    Moving to a Dell VRTX with 2 nodes...

    Exactly, one system to fail instead of two.



  • @garak0410 said in New Server Migration Adventure:

    ...not that makes it easier to migrate off of that one day but very glad to have it.

    Your glad to have a VRTX inverted pyramid? Why? I'd think you'd be livid that someone bought that for your environment. Why are you not upset?

    http://www.smbitjournal.com/2015/11/understanding-the-role-of-the-dell-vrtx/

    Edit: Livid within the scope of understanding it isn't your money or your risk, so not really your problem. But still, not happy. The VRTX is fun to work with, so that's a positive regardless of its applicability to the SMB space.

    We actually had a VRTX thread in ML's first week.

    And a recap last year.



  • I assume this for your main infrastructure. A VRTX in a ROBO for capacity needs is different, where reliability isn't important. The VRTX has a bad track record of being unreliable even within the scope of being an unreliable inverted pyramid design.



  • @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    ...not that makes it easier to migrate off of that one day but very glad to have it.

    Your glad to have a VRTX inverted pyramid? Why? I'd think you'd be livid that someone bought that for your environment. Why are you not upset?

    http://www.smbitjournal.com/2015/11/understanding-the-role-of-the-dell-vrtx/

    I'd have to be upset at myself. I made the final decision after months of negotiating on cost and features. I really tried for a good R730 with Azure as our DR but they didn't want to pay the yearly license. Add this to the Dev projects they push me on, I had had to juggle many things but also try to make the best decision I could.



  • @garak0410 said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    ...not that makes it easier to migrate off of that one day but very glad to have it.

    Your glad to have a VRTX inverted pyramid? Why? I'd think you'd be livid that someone bought that for your environment. Why are you not upset?

    http://www.smbitjournal.com/2015/11/understanding-the-role-of-the-dell-vrtx/

    I'd have to be upset at myself. I made the final decision after months of negotiating on cost and features. I really tried for a good R730 with Azure as our DR but they didn't want to pay the yearly license.

    Why did you compare a VRTX with no DR against a cloud DR alone? Why not just a R730, or two R730? That's not an apples to oranges comparison with the single device VRTX.

    Any why Azure for DR? Azure is, even within the cloud space, extremely expensive. You can do DR for a fraction of the cost without that annual fee.

    But given that the result is not to have DR, the proper comparison would have been other "non-DR" solutions.



  • @garak0410 said in New Server Migration Adventure:

    Thanks for all of the tips thus far. I may have to rush this migration now...

    I came into the office today with the oldest of the two physical servers sounding like jet airplanes taking off..before, it just sounded like a sick car. It could be days our hours before it may give it up. Not sure why the fans are running so high today but add that to some bad memory, failed drives (but still working RAID set)...it simply can't wait too much longer. I asked for this upgrade 6 months ago. But I digress.

    I'd still like to take time and do the migration to 2016 rather than just moving the VM's...luckily the weekend is here (though I have to coach a few ball games this weekend.) But for sure don't want to rush this.

    You are making a huge mistake by mixing the tasks. You new hardware can run everything twice or more, I would assume.

    So just migrate the VMs to stable (I won't call it good) hardware that will not up and suddenly die (most likely).

    Then, next weekend, worry about building Server 2016 and migrating roles.



  • @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    ...not that makes it easier to migrate off of that one day but very glad to have it.

    Your glad to have a VRTX inverted pyramid? Why? I'd think you'd be livid that someone bought that for your environment. Why are you not upset?

    http://www.smbitjournal.com/2015/11/understanding-the-role-of-the-dell-vrtx/

    I'd have to be upset at myself. I made the final decision after months of negotiating on cost and features. I really tried for a good R730 with Azure as our DR but they didn't want to pay the yearly license.

    Why did you compare a VRTX with no DR against a cloud DR alone? Why not just a R730, or two R730? That's not an apples to oranges comparison with the single device VRTX.

    Any why Azure for DR? Azure is, even within the cloud space, extremely expensive. You can do DR for a fraction of the cost without that annual fee.

    But given that the result is not to have DR, the proper comparison would have been other "non-DR" solutions.

    Stop beating a dead horse.



  • @jaredbusch said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    ...not that makes it easier to migrate off of that one day but very glad to have it.

    Your glad to have a VRTX inverted pyramid? Why? I'd think you'd be livid that someone bought that for your environment. Why are you not upset?

    http://www.smbitjournal.com/2015/11/understanding-the-role-of-the-dell-vrtx/

    I'd have to be upset at myself. I made the final decision after months of negotiating on cost and features. I really tried for a good R730 with Azure as our DR but they didn't want to pay the yearly license.

    Why did you compare a VRTX with no DR against a cloud DR alone? Why not just a R730, or two R730? That's not an apples to oranges comparison with the single device VRTX.

    Any why Azure for DR? Azure is, even within the cloud space, extremely expensive. You can do DR for a fraction of the cost without that annual fee.

    But given that the result is not to have DR, the proper comparison would have been other "non-DR" solutions.

    Stop beating a dead horse.

    It's not dead if he was unaware of it. If it was a dead horse, he'd have been aware of the issues.



  • @garak0410 said in New Server Migration Adventure:

    Add this to the Dev projects they push me on, I had had to juggle many things but also try to make the best decision I could.

    I realize this is now a post mortem situation, but I'd say that this is a really important time to take a moment and dig into decision making processes as this is, like any port mortem, an important learning exercise to look at how the process was done, what could be improved, where things went wrong, etc.

    Some thoughts:

    1. You posted about having bought a VRTX and how to use it, but didn't post about considering buying one. We'd have stopped you instantly as it is a known "never for use in the SMB" product that has been heavily vetted for years. We could have sped up your decision process around that from months to seconds. It has been covered, as Jared pointed out, quite a lot. We have lots and lots of resources about it ready to go, both research and discussions. And in multiple communities.
    2. How did a VRTX or an IPOD get on your consideration list, why was it on the short list?
    3. Why did totally different solutions get put on the short list, but not normal ones (like hyperconvergence, stand alone servers, etc.?) While common solutions aren't for everyone, the standard approaches are generally worth at least a brief consideration if only as a comparison point or a "stick in the ground" so that you have a cost and complexity guideline to work against. In this case, you had two extremely disparate and non-standard solution paths that were very difficult to compare directly, and none of the ones more similar to either to see if those were good representatives of their "style".


  • @scottalanmiller said in New Server Migration Adventure:

    @jaredbusch said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    ...not that makes it easier to migrate off of that one day but very glad to have it.

    Your glad to have a VRTX inverted pyramid? Why? I'd think you'd be livid that someone bought that for your environment. Why are you not upset?

    http://www.smbitjournal.com/2015/11/understanding-the-role-of-the-dell-vrtx/

    I'd have to be upset at myself. I made the final decision after months of negotiating on cost and features. I really tried for a good R730 with Azure as our DR but they didn't want to pay the yearly license.

    Why did you compare a VRTX with no DR against a cloud DR alone? Why not just a R730, or two R730? That's not an apples to oranges comparison with the single device VRTX.

    Any why Azure for DR? Azure is, even within the cloud space, extremely expensive. You can do DR for a fraction of the cost without that annual fee.

    But given that the result is not to have DR, the proper comparison would have been other "non-DR" solutions.

    Stop beating a dead horse.

    It's not dead if he was unaware of it. If it was a dead horse, he'd have been aware of the issues.

    The purchase is over and done. It is a dead horse.



  • @jaredbusch said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    Thanks for all of the tips thus far. I may have to rush this migration now...

    I came into the office today with the oldest of the two physical servers sounding like jet airplanes taking off..before, it just sounded like a sick car. It could be days our hours before it may give it up. Not sure why the fans are running so high today but add that to some bad memory, failed drives (but still working RAID set)...it simply can't wait too much longer. I asked for this upgrade 6 months ago. But I digress.

    I'd still like to take time and do the migration to 2016 rather than just moving the VM's...luckily the weekend is here (though I have to coach a few ball games this weekend.) But for sure don't want to rush this.

    You are making a huge mistake by mixing the tasks. You new hardware can run everything twice or more, I would assume.

    So just migrate the VMs to stable (I won't call it good) hardware that will not up and suddenly die (most likely).

    Then, next weekend, worry about building Server 2016 and migrating roles.

    Yes, given the new urgency, just migrate the workloads directly. Upgrade the OS level later.



  • @jaredbusch said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @jaredbusch said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    ...not that makes it easier to migrate off of that one day but very glad to have it.

    Your glad to have a VRTX inverted pyramid? Why? I'd think you'd be livid that someone bought that for your environment. Why are you not upset?

    http://www.smbitjournal.com/2015/11/understanding-the-role-of-the-dell-vrtx/

    I'd have to be upset at myself. I made the final decision after months of negotiating on cost and features. I really tried for a good R730 with Azure as our DR but they didn't want to pay the yearly license.

    Why did you compare a VRTX with no DR against a cloud DR alone? Why not just a R730, or two R730? That's not an apples to oranges comparison with the single device VRTX.

    Any why Azure for DR? Azure is, even within the cloud space, extremely expensive. You can do DR for a fraction of the cost without that annual fee.

    But given that the result is not to have DR, the proper comparison would have been other "non-DR" solutions.

    Stop beating a dead horse.

    It's not dead if he was unaware of it. If it was a dead horse, he'd have been aware of the issues.

    The purchase is over and done. It is a dead horse.

    Ah, that's very bad in IT. You never, ever skip learning from your decision processes. If you treat existing decisions as "over and done" that's how you end up not growing, and repeating those mistakes - where the mistake is the flawed decision making process. Few things are as important, especially in IT, as evaluating the decision process itself. What was missed, how did the decision end up looking viable, what resources were used or missed, etc.

    This is vastly more important than the actual decision. Physical decisions are made once. Good or bad, it's a roll of the dice and that's over. But the decision process is what our careers are built on and something we do over and over again. That's where we grow and how we get better at what we do. In IT especially, we can only ever know so much, but we can improve the process by which we leverage knowledge greatly.

    It's probably the most important thing that we do in IT, constantly going back with a critical eye to how we make decisions, and improving that process every time. Never skip that evaluation, that's 100x more important than the individual decisions.



  • @scottalanmiller said in New Server Migration Adventure:

    @jaredbusch said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @jaredbusch said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    @scottalanmiller said in New Server Migration Adventure:

    @garak0410 said in New Server Migration Adventure:

    ...not that makes it easier to migrate off of that one day but very glad to have it.

    Your glad to have a VRTX inverted pyramid? Why? I'd think you'd be livid that someone bought that for your environment. Why are you not upset?

    http://www.smbitjournal.com/2015/11/understanding-the-role-of-the-dell-vrtx/

    I'd have to be upset at myself. I made the final decision after months of negotiating on cost and features. I really tried for a good R730 with Azure as our DR but they didn't want to pay the yearly license.

    Why did you compare a VRTX with no DR against a cloud DR alone? Why not just a R730, or two R730? That's not an apples to oranges comparison with the single device VRTX.

    Any why Azure for DR? Azure is, even within the cloud space, extremely expensive. You can do DR for a fraction of the cost without that annual fee.

    But given that the result is not to have DR, the proper comparison would have been other "non-DR" solutions.

    Stop beating a dead horse.

    It's not dead if he was unaware of it. If it was a dead horse, he'd have been aware of the issues.

    The purchase is over and done. It is a dead horse.

    Ah, that's very bad in IT. You never, ever skip learning from your decision processes. If you treat existing decisions as "over and done" that's how you end up not growing, and repeating those mistakes - where the mistake is the flawed decision making process. Few things are as important, especially in IT, as evaluating the decision process itself. What was missed, how did the decision end up looking viable, what resources were used or missed, etc.

    This is vastly more important than the actual decision. Physical decisions are made once. Good or bad, it's a roll of the dice and that's over. But the decision process is what our careers are built on and something we do over and over again. That's where we grow and how we get better at what we do. In IT especially, we can only ever know so much, but we can improve the process by which we leverage knowledge greatly.

    It's probably the most important thing that we do in IT, constantly going back with a critical eye to how we make decisions, and improving that process every time. Never skip that evaluation, that's 100x more important than the individual decisions.

    A postmortem of a process is a great tool to learn about your decisions. But you were beating this into the ground as if it was an active decision. It was not until your previous post that you began to treat it like a postmortem analysis.


Log in to reply