Thoughts about hosting

Started by JPinto, Nov 11, 2022, 01:15 AM

Previous topic - Next topic

JPintoTopic starter

Hello. I am interested in the opportunity to raise for my needs (there are a couple of projects that I would like to launch, but resources are required that are expensive if you buy from scratch and somewhere on the side, and there is your own unattended hardware) web hosting. Can you tell me if the right course of thought about infrastructure?
 And in general, I would like to hear from people who are in the subject of some good advice.
Infrastructure: 1) the router provides internet access. 2) a virtualization server with a hypervisor (I give 70% of resources to the main task, the remaining 30% will go to another pair of virtual machines for other purposes unrelated to hosting). 3) network data storage (large-volume storage, for storing virtual disks of machines).

I put the OS on the hypervisor VM, on which I roll the billing + hosting panel. Billing is needed because there are a couple of people who ask to share resources - I want to have automation of processes so as not to think, not to guess and not to follow with a pen, notebook and calculator.

1. Tell me a budget and / or free bundle of billing + hosting? Of course, with the possibility to Russify yourself.
2. Do I think correctly that I need a large disk inside the virtual machine that is stored on the storage and as I use it up, I increase its size and add disks to the storage? Or you need to create only a system disk, and throw a folder into the hosting OS (well, or by iSCSI) Storage - where everything will flow. Which option is better, more stable, scalable, more reliable?
3. The question is unclear how to connect an external white IP address to the virtual hosting of a specific user? I need 2-3 addresses to connect to 2-3 hosting accounts - how does it work at all? Which router should I have? How is it prescribed in general and where is the most interesting? Here the provider gave me the addresses of 5 white pieces, for example, and what to do with them how to connect them and use them? For example, I put a static address in the router and forward port 80 to the address of the hosting panel, so in the router it is possible to connect only one address. In a word, a dark forest.
4. Let's say there is server # 1 , virtualization where billing + hosting, everything works. I want to add a server to transfer some account to another server, what should I do? I put the OS + hosting panel on server No. 2, and add it as resources in the billing on server No. 1?
5. As I understand it, it is desirable to bring the billing panel to a separate server? Is there a weak server on which you can do this - will it pull? I mean, is there any load on the billing? Or is there always a big load on the billing and hosting panels without a difference? I don't understand how the traffic is being chased in the locale.
6. What to do with the courses in the locale? Let's say there is a network drive in the locale visible in the locale with its name. If I allocate any resource in the virtual machine, it turns out that he sees it over the network, which is not good - how to isolate virtual machines from local resources? If you give other subnets to virtual machines and hosting, then how are external white IP addresses sent?

Everything is clear with the hypervisor, I chose Proxmox. The rest I don't even know that I didn't stumble on the network - it's leaky, it's unclear with localization, then billing doesn't fit the hosting panel, etc. Direct in the right direction, comrades,
I really want to touch my web hosting with my hands, and the iron is gathering dust and the server cabinet is disassembled waiting in the wings.


You probably have an Ethernet cable at the input without any PPPoE authorization?
The router is not needed here.

Just plug the provider's cable into the hypervisor machine.
And you distribute white addresses to virtual machines.

It is unlikely that you have a cool storage system.
I would generally rely on the local disk of the computer. This is more reliable than a cheap storage (the one that is cheaper than $1K without disks).
You don't have a cloud storage system, do you?
So what kind of scalability are we talking about?
A separate storage system will make it easier to replace the server hardware without transferring disks (data on disks). There is no more benefit from it, except harm (an extra entity in the system is an extra point of failure).


the biggest surprise for me was the blatant style of work of the Swedish provider — . I expected anything but the absolute helplessness demonstrated by one of the largest European hosts. The struggle to provide for basic needs lasted four days — God knows, I endured to the last! — and ended in a complete collapse. None of the assigned tasks it was not solved — mail redirection via Spamorez could not be adjusted, the mail functions in the scripts of my site never worked, access to the logs was not provided, the mystical traffic leak continued until the last day.

It is not the fact of the occurrence of problems that is tragic, but the inability or unwillingness of the so-called support service to fix these problems. Absolutely outrageous: as soon as Stephans, Karl and Pauli realized that it would not work to get away with primitive phrases from the FAQ, they lay low, playing a vulgar silence.

After a spin on the host review, I finally became convinced that I was not on my way with domestic hosting, so I switched to , in order to explore the offers on the American market. The top ten most reliable hosts were headed by a certain — that's where I started. I missed the phrase about 99.9% of uptime because of its standardness, but I was inspired by information about the policy of unconditional refund of money for 45 days in case of dissatisfaction with the service, as well as statistics on the provision of services to more than one million domains.

The last figure seemed to me not quite unambiguous: a million is, of course, wonderful, but what kind of technical support service do you need to have in order to satisfy the requests of such a mass of users? According to , support is provided by 150 employees, it turns out — one for 7 thousand domains. N-yes... An unexpected surprise turned out to be the hosting provider's forum, where everyone shares their impressions about the work . Needless to say, this is a bold unprecedented step, especially since I did not find any censorship on the forum: the most unflattering statements by the company's employees were not illustrated, but commented on, and with a high degree of persuasiveness.

The offer itself it was amazing: for $ 14, 95 cents, the client gets 1 terabyte (!!!) disk space  plus... unlimited EVERYTHING: traffic (au, !), the number of domains and subdomains for servicing ftp accounts, MySQL databases, mailboxes. I confess, I have never seen a package of hosting services with 69 options in my life! Don't be lazy, take a look for yourself, it's just incredible.

Well — I go for it! Five minutes to fill out the form and specify the credit card number — and a letter falls into the mailbox with a notification that everything is ready to work. Having dropped the contents of the site to a new location via ftp, I created mailboxes, ftp accounts and started setting up scripts. The setup, however, did not take place, because there was nothing to configure: every single script earned right away without the slightest additional fine-tuning! Of course, the standard configuration of the Linux server affected, nevertheless, this is a clear merit , which kept this configuration unchanged without succumbing to proprietary temptation.

Considering the control panel (cPanel X) unprecedented in the number of settings , I realized that the phrase the objective need to use a proprietary control panel due to the solid volume of the customer base is a crafty excuse. The only purpose of proprietary panels is to limit the flexibility of site management on the part of the user, and not to create additional amenities for him. Bullshit!

Anyway, within an hour the Virtual College website was in full working order. Without a single request to the technical support service. This circumstance immediately explained the paradox with the maintenance of a million domains by 150 employees: it is enough to correctly configure the server database, and the frequency of contacting the support service will decrease significantly.

The records changed, the mail went to Spam, and did not come back in any way — Hostgator refused to accept it, requiring authentication, and this is wrong, because authentication is used when communicating mail "server-client", and not "server-server".

I contacted the support service at three o'clock in the morning, local — American— time. The answer came... in eight minutes!!!

I will not draw a splint: it took exactly 33 hours to resolve the problem with the Spam cutter, and an intensive discussion took place both day and night — the service worked non-stop. It worked, not imitated the activity. She worked, demonstrating brilliant professionalism, since even the guru from Spamorez identified the configuration deadlock as almost hopeless.

After numerous unsuccessful attempts to arrange the forwarding of filtered mail back to the server it seemed that the matter was in the fundamental security policy of the American hosting provider. As Alexey Cherepov suggested: "It is quite likely that the support service will not be able to establish access at all. The issue here is more political than technical. Like, if you change MX records, then you drive by mail yourself, we won't take it back."

Meanwhile, there was nothing to do without Spamming on your own domain: every hour about three hundred letters with advertising garbage arrived, which, although marked with Spam Assassin in the header ***SPAM***, however, required an incredible amount of time for purely physical removal of letters from the server. The situation was becoming critical, and Alexey Skerepov put forward three ways out of the situation: 1) change the provider again, 2) place mailboxes on the Spam server, 3) seek technical support reconfiguring servers so that they eventually skip mail with someone else's MX records.

I chose the latter option, especially since the work of the Gator support service was encouraging in its dynamics. Unlike who, after a serious difficulty, stopped responding to letters, the Americans did not drop out of contact for a moment and constantly tried to change something by virtue of their competence. It was funny to observe how the qualification of the personnel connected to solving the client's problem increases from time to time:
Mail, a first-level system administrator, came to the aid of the midnight cowboy Herbert Jones, who suggested that to forward mail from the Spam server, the EHLO (Enhanced Helo) command should be used instead of the traditional HELO. We tried it, but it didn't work out. Karl, the administrator of Linux systems, who had been setting up something for a long time and also to no avail, connected.

It seems that the Americans themselves were turned on in earnest, accepting the challenge of a malicious penguin. I can see it right with my own eyes, as in the corridors of the office a thunderous cry rang out: "CALL VIYA!!!" And Viy appeared — Jeremy, the administrator of Linux systems of the first level: "I know what the problem is," the demonic spirit immediately cut off. — When adding an external MX record, our server automatically registers it in the remote domain file, whereas in your situation an entry in the local domain file is required.

A minute later, the MX record was changed, and a minute later, Jeremy manually made the necessary corrections on the server , and half an hour later (after the deployment of the domain name), the mail filtered by the Spam Cutter floated into my mailboxes. The issue has been resolved — the last link in the site configuration has closed after the move, and, hopefully, I will have at least one headache less for years to come.

Well, what can I say? That's how you need to work, gentlemen-fellow hosts! It's not even about the highest qualifications and competence, which, it seems, cannot be found in the technical support services during the day with fire (and after all, Jeremy is "just" a first—level administrator, there is also a second one!), but in that very final phrase: "Thank you for your time and have a great day!"