Română Dedicated servers
Physical machines in our data centre, rented whole. This page covers how one reaches you, the few things you can do with it yourself, and the longer list that needs a person.
What it is
A dedicated server is a physical machine in our data centre, allocated to one customer. It is not a virtual machine: there is no hypervisor between you and the hardware, and one machine carries exactly one service. The platform tracks the actual chassis, including the rack it stands in, its position in that rack, the switch and switch port it is cabled to, and the MAC address of its network interface. The operating system is installed onto the disks in that chassis, using the partitioning and RAID configuration from your order. That difference is the reason this page reads differently from the cloud server page: work a hypervisor does in a second on a virtual machine here needs someone standing in the data centre.
Getting started
- Choose a configuration on the dedicated servers page. The order form starts from a template and lets you change the number of processor sockets, the memory, the disks, the RAID configuration, the network card and the traffic allowance.
- Some machines are already built and are offered at a fixed price. Ordering one of those skips the build step below, because the hardware already exists.
- Place the order and pay it. Provisioning starts when the payment is confirmed, not when the order is placed.
- When the payment is confirmed, the platform looks for a machine that is already racked, marked available, and matches your order on processor, RAID controller, memory, disks and extra hardware. If one matches, it is assigned to your service.
- If nothing matches, the machine has to be built and cabled first. That step is manual: the rack, the position in the rack, the switch port and the MAC address are entered by our staff, and provisioning cannot continue until that record exists.
- The operating system is then installed over the network onto the machine's own disks, with the partitioning and RAID from your order, and the machine is powered on.
- Log in with the username and password shown on the server page in the panel, under Server Details. The account is root unless we tell you otherwise, and the password is generated when the service is created. There is no SSH key step for dedicated servers.
Provisioning is not automatic and it is not instant. We do not quote a turnaround here, because nothing in the platform measures one: if a machine matching your configuration is already racked it is assigned to you, and if it is not, a person has to build it first. Ask support where your machine is and you will get a real answer rather than a guess.
Networking
Each server is cabled to one port on one switch in one of our racks. That port is configured at either 100 or 1000 Mbit, depending on the network card in your configuration.
The port also carries an access list that permits only the IPv4 addresses assigned to your service and denies everything else. Adding an address to the interface inside the operating system will not make it work: the address has to be allocated to the service first, or the switch drops the traffic.
Addresses come from a VLAN reserved for dedicated servers. The VLAN carries the gateway, the prefix length and the resolvers, and those values are written into the operating system when it is installed. Addressing is static, not DHCP, so if you reconfigure the interface by hand you have to keep the same address, gateway and netmask.
Your configuration comes with a number of included IPv4 addresses. Up to that number you allocate them yourself, from the IP Addresses card on the server page. Once you are at the included count the button is gone and the panel tells you to open a ticket.
Reverse DNS is yours to set, one name per address, from the panel or through the API. The name has to contain a dot and must not start with one. When an address goes back to the pool its reverse DNS is reset to hostedby.pidginhost.net.
IPv6 is a different story, and the honest version is this. The platform models IPv6 for dedicated servers, the VLANs carry an IPv6 gateway, prefix and resolvers, and a server can hold IPv6 addresses. What does not exist is a self-service path. Provisioning allocates IPv4 only, the address table in the panel lists IPv4 only, and the API returns IPv4 only. An IPv6 address on a dedicated server is assigned by support.
Limits
These come from your own server configuration or from the platform, not from a per-account quota. Where support can change something, it is changed by support and not from the panel.
| Limit | Value | Can support lift it? |
|---|---|---|
| IPv4 addresses you can allocate yourself | The number included with your configuration | Yes, on request. Your configuration also carries a maximum |
| IPv6 addresses | None are allocated, and there is no control for them | Yes, support assigns them |
| Operating systems offered for reinstall | The images enabled for dedicated servers | Yes, ask for an image that is not in the list |
| Power actions and reinstall | Only while the service is active | No. The service has to be active first |
| Controls while the machine is in maintenance | All of them are disabled | Support sets and clears maintenance |
| Hardware changes after provisioning | No self-service path exists | Support only. It is a physical change |
| Reverse DNS name | Must contain a dot and must not start with one | No. It is a format rule |
| Opening a server page as a sub-user | Needs the credential reveal permission, which read-only members do not have | Yes, the account owner grants it |
One thing the table cannot express. The power state shown on the server page is the last state the platform set, not a live reading from the machine, and the job that would reconcile the two on a schedule is switched off. If you shut the server down from inside the operating system, the panel can go on showing it as running.
What needs support
This is the long section on this page, and that is the point. On a cloud server almost everything is a button. On a dedicated server it is not, so here is the whole list of what has no control in the panel and no endpoint in the API.
- Getting a machine racked when your configuration is not already in stock. There is no automated build, and provisioning stops until the hardware record exists.
- IPv4 addresses beyond the number included with your configuration.
- Any IPv6 address at all.
- IPMI, or any other out-of-band console. The platform stores IPMI credentials against the machine, but only staff can see them. There is no panel page and no API endpoint that hands them to you.
- Booting rescue media, or anything other than the installed operating system. The only install path exposed to you is a full reinstall, which runs the network install again from your order.
- Hardware changes: processor, memory, disks, RAID controller or network card. The modify action on a dedicated service does nothing but notify us, so treat it as a ticket.
- Changing the login name, or resetting the password without a reinstall. The panel reveals the stored password, it does not change it.
- Changing the hostname on the service after the order.
- Putting a machine into maintenance, or taking it out again.
What you can drive yourself is the short list: start, soft stop and force stop from the panel, the same four power actions plus restart through the API, a reinstall onto one of the offered images, allocating an included IPv4 address, and setting reverse DNS. If an operation is not on that list, assume it is a ticket.
Where to go next
Something here still not clear?
If a limit on this page surprised you, tell us. That is a documentation bug and we fix it.