Answer the question
In order to leave comments, you need to log in
How to correctly transfer the entire infrastructure to virtualization?
Good afternoon, ladies and gentlemen.
I tried in vain to figure out the current situation myself, but I feel that without a collective mind and without experience it will not work.
The question will be big, at least it seems so, and versatile. I'll start, I guess.
How does everything work now?
There are three HP ProLiant DL360e Gen8 servers:
1. E5-2403, RAM 3x4, HDD 2x1Tb
2. E5-2403, RAM 2x4, HDD 2x4Tb
3. E5-2403, RAM 2x4, HDD 2x4Tb
On all Windows Server 2012 servers.
On the first DC, DNS, DHCP roles; the second as a file server; on the third all backups of the first two.
There are two more servers assembled on the knee, one in the role of a proxy server (Kerio Control), the second with the Guarantor / Consultant stray based on Windows Server.
The network has about 80 users, an MFP, a couple of Wi-Fi points.
There are no routers, there are three switches.
What would you like ideally?
I still want to virtualize the infrastructure. I study VMware vSphere, I tried to install it, I liked it very much.
We plan to purchase storage systems, one for now, in the future one more to make sure.
Probably you need to buy a router in order to scatter servers and users over vlans.
I know that I can install vpshere on all servers, assemble them into a cluster, and raise virtual machines.
All in all, a bit of a mess. How to transfer the file server so as not to edit the rights again?
Will it be possible to forward a network card or even two so that the proxy server is also a virtual machine?
How to connect three servers to storage? Is it possible to make two logical partitions on storage, so that virtual machine files are in one, and backups are added to the second using Veeam, for example.
So far we are thinking of installing some kind of qnap.
Should I get a router?
I don't even know what else to write.
Ask questions, right? :)
Thanks a lot in advance.
Answer the question
In order to leave comments, you need to log in
1) I support speakers who talk about deploying machines from scratch in a virtual machine, without trying to convert P2V. When converting, different artifacts may pop up, well, what for. A well-planned migration will not even require service downtime.
2) Give storage only for virtualization, no backups there. In your case, it's like shooting sparrows from a cannon. For backups - pull out all the screws from your proliants and put them on one of some servers, and make a backup storage based on it. ESX can boot both via iSCSI and from a flash drive. The proliants have a compartment for an SD card, you can stick a flash drive with ESX in there. In other words, hard disks are not needed for virtualization hosts.
3) Kerio - fu-fu-fu! :-) You don't need cards, a hardware router, in principle, too. Each provider in a separate VLAN, both VLANs - in a virtual machine, on a virtual server with FreeBSD, for example. And that's it, route as you like.
4) QNAP as a storage system - fu-fu-fu! Some kind of self-assembly is better (a server with a large disk basket, or a separate pair of servers + a DAS basket), with FreeBSD and ZFS inside, and even with Windows. It's cheaper and easier to maintain. NAS-s from the cheap segment have artifacts that are very diverse and very bizarre. LUNs fall off, rights fly off, generally disappear from the network. In short, they got drunk, it was the case.
5) Think towards Windows. Please note that Windows 2012 R2 Standard as a virtualization host (Hyper-V) allows you to virtualize 2 Windows servers internally under the same license. In other words, if you have 3 licensed Win2012 R2 hosts, then with their help you can virtualize 6 servers with Windows without buying any additional. licenses.
Hyper-V can run virtual machines directly on SMB 3.0 file balls. That is, iSCSI, FC and other fashionable technologies from the NAS / SAN area are not needed. Enough Win2012 R2 and an open file ball on it. Windows knows the Storage Spaces technology. Which (technology) can even automatically tiering, and out of the box. QNAP, which can do this, will cost thousands under 300 rubles. This is without discs.
Based on the set of services that you have on the network, you, in principle, do not need storage. Not those speeds, not those volumes. You do not have 10-gigabit links, do I understand correctly? Look instead of storage at some servers with large baskets. Well, let's say from 8 to 24 disks. Almost everyone has 8, Dell has 10 and 20 disks, STSS has servers with a 24-disk bin ( this is what it looks like). Or on DAS (direct attached storage). You need a server (almost any), a SAS-HBA server with an adapter with a pair of external ports, and a DAS basket that connects to this adapter with SAS cables.
Also take into account that branded storage systems (even QNAP) are a thing in themselves, and it is quite difficult to diagnose there. Therefore, support is usually bought from the vendor. And it also costs a lot of money. And without support, climbing into the giblets of the storage system yourself is fraught with a rake of this size. Self-assembly is easier to diagnose and easier to repair (change components), if suddenly something happens.
everything is possible. but it is not recommended to transfer realki to a virtual machine. something will always come out. just move the roles and transfer the data itself. instead of virtualizing machines.
1. Iron machines can be packed into special virtual machines. the same vmware, VMware vCenter Converter
2 has software. (without forwarding).
Are you seriously? Store the datastore in the storage system and backups in the same place? And the storage system will fly, what will you do?
You can migrate Windows systems to a hot one using the Disk2vhd utility and you won’t need to reconfigure anything, just turn off the physical servers and start the virtual one from the image. There is a successful experience of migrating to a hot one using this utility, but the truth is that they launched virtual machines under Hyper-V.
Why do you need a sphere when Windows has Hyper-V? By the way, I remember MS does not recommend to verticalize AD.
By proxy, by vlans, throw network connections through a distributed switch in the vmvar and then resolve it on the switches. And about storage, don't keep all your eggs in one basket, make backups on a separate backup server, because you don't want to break all the eggs when the basket falls. The domain can theoretically be steered on a virtual machine, but how will you solve the problem with the clock? And the amendment is not a router, but a switch is needed to throw users over vlans.
With such volumes, the virtualization idea is controversial. Is it necessary?
If so, what will be added to it? You do not have such a large fleet of everything to justify the cost of virtualization.
From councils:
Don't even try to look towards vMware. The system is great and works well. But, as soon as you leave one server, the prices are very biting. The whole point is that without a Management server, it is very inconvenient to manage everything. Clustering is only available with Essentials Plus (~$4500) and that's for three two-socket servers. Switching to Standard is about $16,000 - $18,000. All cheerfulness is understood as soon as needs increase.
Therefore, it is better to immediately look towards Citrix XEN server or Proxmox VE.
I would advise you to simply raise the Hyper V role on the third server and, as sharkirill said, regularly (every ....) run disk2vhd for hot conversion. In the event of a physical server failure, you can quickly start working again.
But if you really do something useful - it's better to virtualize workstations. With such a number of workers. places, the economic effect is VERY noticeable. Especially as the park ages. Yes, and the administration of the VDI park, time is spent many times less.
My most important advice is not to listen to active people in this all "five times a day ..."
The system administrator must be smart and lazy, that is, automate everything and not even control some strange operations :-))
1 ) The sphere is really paid, a simple edition costs about 900 bucks. But the license includes three servers and a vcenter server. Great option for you.
2) Even this version of the sphere includes a lot of interesting things that Hyper-V definitely doesn’t have
3) Regarding storage, the ideal option from vmware is VSAN or VSA ... But this is really expensive. Including for hardware (a lot of disks and normal controllers) - but an important plus is that you don’t have to buy some dubious Chinese crafts, not even at the SOHO level, not to mention the enterprise
In terms of price, these solutions are positioned as an inexpensive replacement for a full-fledged SAN.
But personally, with a limited budget, I did it easier and assembled my own iscsi-target cluster - two pieces of iron, two controllers, two sas candles, and two HP D2700. Budget MPIO has been released with high avail.
4) One processor? E-series? Well, this is not serious :-) In fact, this is certainly enough for an average office. I have been running a VSA cluster on 3xML150 for two years in one such office, and about 30VM
5) Everything can be virtualized. A domain controller is no exception, another question is that if something goes wrong, then you need to restore it very carefully. Although if there is only one controller and there is a regular backup of it entirely, then everything is ok. Complex applications - with caution, but the same is possible and necessary.
6) In 2006, I had a domain controller under VMWare Workstation on my laptop for a whole week and served a grid of 600 machines - there was a failure, I had to get out like this, I know that this is the microsoft way, but ... I think about the issue of performance answered :-)
7) And P2V conversion - no problem, BUT in the same vmware, for better performance, their own disk controller drivers (VMWare paravirtual) are recommended, and this is not so simple when converting Windows machines.
Didn't find what you were looking for?
Ask your questionAsk a Question
731 491 924 answers to any question