Dedik one should be taken or how to make fault tolerance?

Question

Network administration

Now the project is on VPS. But we don't have enough power. And some of the team members are drowning for the purchase of a VDS tariff. dedicated server. But in this case, if we buy a Dedik, then all the elimination of problems with the server will fall on us, if before that the support somehow helped. Plus, if their vps dies now, they will restore all the information. And with the move to the Dedik, if the SSD burns there, then no one will be able to restore the information to us. I'll have to take it from the archives. The site will have a simple.

What we want is an uninterrupted operation of the site with the guarantee that if one database goes down, then its mirror copy will connect. If the web server goes down, then a second copy from another server will connect. if the disk fails, the site will not lose its functionality.

What technologies can achieve this? Take two dedicators and set up mirroring or what?

And the project is a trading platform where there are many vocrets that update prices, product lists every 5 minutes and orders are placed. I believe that such a site should not be allowed to be idle for more than a minute. what solution to choose?

So far, vps, but there are also vps for storing static data and for development devs. But now money will be invested in the surge of visitors, and the market is limited, and if they go to the site, and the site is not working, then they may no longer come, and this is important. after all, in this business segment, the number of customers (target audience) is measured in hundreds, not thousands.

Answer 1

What answer do you want? What is redundancy and balancing by your forces? Yes it is. Why is it necessary? That you will deploy your virtualization, your DNS, balancers, your databases with replications on dedicated servers? Yes it is. What will be better? Well, if you understand that you need your own virtualization (and you are ready to organize it financially), then yes, you need your own dedicated server. If you don’t understand this yet, then VDS is not needed.

Answer 2

The most important thing to understand is that VDS is also a virtual machine like VPS.
It is not entirely clear why you need to move to a dedicated server. Can't your vps/vds provider give you the right config? Usually they can give even more than most dedicated servers.
Fault tolerance is already in the clouds you need to scratch. There it is as simple and convenient as possible. If you need a rashka - Yandex cloud, selectel, mail ru and others. The first two have the most developed infrastructure.
There are no special problems to scale web servers. You put nginx in balancer mode and balance traffic on the web server.
Tin begins with the base scale. Managed service will help here, for example from Yandex. There it is proposed to immediately create a cluster of three servers in different data centers. This simultaneously allows you to distribute the load and implement fault tolerance.
But I say right now. It's not very cheap. How expensive? Protest - you know. It's hard to say something, because there are no figures from you on the load.
Of course, you can do all this on ordinary vds / vps. But you will have to fool around if you don’t know what and how to do, mom don’t worry.

Answer 3

What technologies can achieve this?

The technology of hiring a normal admin to the team.
Obtaining fault tolerance and reducing the recovery time in case of failures is a whole range of measures, just a conditional "cluster of servers" or "transfer to the cloud" cannot be achieved, no matter how the apologists for the orchestration of kubernetis sing to you - sooner or later, ears of unforeseen circumstances will start to stick out from somewhere .

Answer 4

Fault tolerance can be done indefinitely. Make backups of the entire VDS and don't fool around.
If everything is serious there, then be so kind as to share the money with a competent administrator.

Answer 5

1. You do not need a cluster of many servers yet, since the project lives on vps and you are just starting to think in terms of increasing power.
2. Take two dedicated servers
3. On the first server, raise a virtual machine (give it 99% of all resources).
4. Deploy the project only in a virtual machine (it only means strictly!)
5. Every night you make a snapshot of the virtual machine and dump it on the second dedicated server.
6. In the morning, when you come to work, deploy the snapshot to the virtual machine on the second server and check (mandatory) that everything works. Now, if something happens to the first server, just switch (to dns if, as I understand it, you have a website) to the ip of the second server.
Cons:
1. It is more expensive to rent two dedicated servers than vps, as you have now.
2. At the time of creating a snapshot, the virtual machine stops (that's why I wrote what to do at night).

Answer 6

I agree with Valentin. In any case, you need a mountain reservation. And how you will achieve it - a bunch of VPS, 2 dediks and virtual machines on them, AWS or something else ... it's up to you