Which solution to choose for creating a budget failover cluster?

Question

Virtualization

Hello!
Help with the choice of what equipment and what software to use to create a budget failover cluster.
Initial data:
One physical server running the ESXi 6 hypervisor.
An organization with 60 terminal users (it is planned to increase to 100), users do not have their own desktop computers, they all have thin clients. The main pattern of user work is a running session with 1C:Enterprise 8 (they work in a configuration specially designed for them) and a huge bunch of tabs in the browser (which currently creates a huge load on the processor). The LibreOffice package is installed on the terminal server, but it is run so rarely that it does not affect performance at all. This is one virtual machine.
1C infobase running 1C:Enterprise server, Microsoft SQL Server 2012 Standard DBMS with Service Pack 2 (SP2). The current size of the database is 9 GB (approximate growth of 3 GB per year). This is the second virtual machine.
There are also about 5-7 virtual machines on the server now that use the minimum amount of resources, but they are needed to maintain the infrastructure (telephony, wireless network controller, monitoring).
Now it all works on a server with the following specifications:
- X9SRi-F motherboard
- Intel Xeon E5-1620V2 processor (4 cores, 10M Cache, 3.70 GHz, DDR3-1866)
- 32 Gb RAM (DDR3-1866)
- RAID controller LSI MegaRAID SAS 9271-8i, 1GB, 8int ports, 6Gb/s
- Intel S3500 Series, 120GB - 2 pcs.
- Hitachi Ultrastar A7K4000, 2TB - 2 pcs.
Question:
I need your advice on organizing the smooth operation of users. Since now everyone works on the same physical server, it is problematic to perform any routine maintenance on it and there are only a few hours a week when it can be serviced, and some non-standard situations generally turn into serious problems. I would be grateful for your advice on what equipment to buy in order to achieve uninterrupted operation and what technologies to use for this. Now the virtual machines are running VMWare ESXi 6.
Are there any budget methods for organizing collaboration between a server and a disk array, or given our volumes and tasks, is it too early to think about this?
I read that storage systems can be assembled on ScaleIO and in this case it will turn out to be both computing resources and a fault-tolerant disk system. What do you think about this option, are there any best practices for building low-cost fault-tolerant systems?

Answer 1

ScaleIO needs at least 3 hosts, otherwise it won't take off. If you are not considering software, so to speak, "filibuster edition", then $10k is not enough for you. Iron can be, for example, this, in an amount of three pieces:
stss.ru/products/servers/T-series/TX111.4-004LH.html
Complete with 32 gigs of memory, 4-core HT processor and 4 x 300 Gb SAS 10k RPM screws for one server piece is 172 thousand rubles. In total, respectively, about 510 thousand. That's all your 10k$.
But on this it is already possible to deploy ScaleIO. If three servers are in exactly the same configuration as I wrote above, then from 3 hosts you will get 1.2 TB of space. Sparsely, but this is a really converged fault-tolerant solution, with the ability to decommission any host (but only one).
Well, this is the price price of STSS, if you cry for a long time, you can convince them to drop the price. The license is still not enough, but a penny is never superfluous :-)

Answer 2

1 buy two identical windows servers of the required capacity
2 install a virtual domain controller on each locally, create a domain, add servers to it
3 install starwind* for each, configure disk export via iscsi, configure synchronization
4 connect disks
5 assemble servers into a cluster
6 add disks as shared disks of the cluster
7 add the role of a failover file server to the cluster, create shared folders
8 move virtual machines to a shared folder on a failover file server
9 wait for synchronization of starwind nodes
all. from now on you have a fully failover and highly available server cluster. you can cut wires on any one server, all virtual machines will instantly move to the second one without shutting down / rebooting and downtime, not even a single "ping" will be lost
* free starwind has limitations, and paid up to 4TB of disk space costs about 150k rubles, which is almost free for this functionality