Virtual Machines

A virtual machine is one running guest on one of your hypervisor nodes. This article covers the list of every machine you manage, the per-machine screen and everything you can do from it, and which of those things a customer can do for themselves.

The machine list

Open VPS in the admin sidebar, then the All VPS tab. The header counts the machines and the clusters they are spread over. The same table, with the same controls, also appears on a single cluster's page showing only that node's machines.

Finding a machine

The search box matches on machine name, address, cluster, service number and customer e-mail. Alongside it, dropdowns filter by cluster, status, plan, tag, origin (created here, or imported from another panel) and service link (linked to a service, or not linked). Clear filters resets them all.

Columns

A Columns button chooses which columns are shown, and the choice is remembered in your browser. Reset to defaults puts it back. Available columns are the machine name, status, vCPU and RAM, disk, IP addresses, primary IP, plan, cluster, customer, service, tags, created date and this month's traffic. Name, status and the actions menu are always shown. Adding the traffic column costs one extra read per cluster, which is why it is not on by default.

Statuses

What each machine status means

Status	Meaning
Creating	The build is under way. The machine does not exist on the hypervisor yet.
Running	Powered on.
Stopped	Powered off. It still holds its memory allocation and its disk as far as capacity is concerned.
Paused	Frozen in place. It keeps its memory but answers nothing until it is resumed.
Migrating	Being moved live to another node.
Error	Something failed. Look at the machine's Activity section for the job that reports why.
Orphaned	The record exists in the panel but no matching machine was found on the hypervisor. Records are never removed automatically — verify on the node first, then delete the record explicitly if it really is gone.
Terminated (disk retained)	A machine whose service was terminated and whose disk is still on the node awaiting a wipe.

Rows also carry markers: Unsold when no billing service is linked, the source panel when a machine was adopted, rescue mode while it is booting from rescue media, and a note when the machine sits on hardware another panel owns. A machine adopted from a panel that manages guests in its own way carries a badge saying a conversion is still pending; converting it costs one power cycle and is described in VPS Panel Migration.

Acting on several machines at once

Tick rows to reveal a bulk bar offering the power actions that apply to every machine selected. The confirmation lists each machine, and afterwards the panel reports each one individually as accepted or refused, with a reason. Select the ones that failed re-selects only those so you can retry. If no single action applies to every machine in its current state, the bar says so.

Export CSV writes every field of every row matching the current filters.

An Auto-refresh switch re-reads the list every 30 seconds while the tab is visible.

Creating a machine by hand

Open the cluster you want the machine on and use Create VM. This builds a machine directly, with no service and no invoice behind it — use it for your own machines and for testing.

Note: to create a machine for a customer, place an order instead: the Create menu on the tab strip, entry VPS, opens the new-order form filtered to VPS products. A machine built by hand has no service, no invoice and no billing cycle until you link it to one.

Fields on the create form

Field	What it does
Plan	Pick a plan to take its resources, or Custom (no plan) to type them. With a plan chosen, Override plan resources lets you change the numbers for this build only.
vCPUs, Memory (MB), Disk (GB)	The size, when there is no plan or you are overriding it.
How many	Build several identical machines in one go. Each gets its own name, and that name is its hostname. The form previews the first and last name and reports progress as they are created.
Storage type / Storage pool	Where the disks are created. If the node has pinned pools, the form says so and offers them; you may still build elsewhere by switching the choice explicitly. If the node has never reported a pool of the type you chose, the form tells you to refresh its inventory or type the name the node uses.
Bridge	Which of the node's bridges the machine's interface attaches to.
Allocate IPs from IPAM	Reserves addresses from the pools serving this cluster's location. With a plan chosen, the form says how many it will take.
OS image	The installer. With a plan chosen, only that plan's catalogue is listed — anything outside it is refused when the job is queued. None — blank disk or custom URL builds an empty disk.
Source image URL	Seed the disk from an image address instead. A hand-typed address carries no checksum — the node writes whatever the other end answers with. Prefer a catalogue image.
Hostname, SSH public keys, Root password	Applied on the machine's first boot. One key per line; a key means the machine is reachable without ever setting a password. For the password, choose Keep the image default or Set one now — there is a Generate button, at least 8 characters is required, and it is never shown again once the form closes, so copy it while it is on screen.

Note: a plan's install script is not run by a manual create — only an ordered machine and a reinstall run it. A machine built here comes up as the bare operating system. Images without first-boot configuration support ignore the hostname, keys and password entirely.

A machine's own page

Click a machine to open it. The header carries its status, its cluster, its plan and its service, with the action menu beside them. Seven sections follow, and the one you open is part of the address in your browser, so you can link a colleague straight to it.

Overview

Specification (vCPUs, memory, disks), operating system, boot order, creation date, the machine's identifiers on the hypervisor, and a 24-hour usage trend. Where the guest reports its own operating system, that is shown alongside the image it was installed from. Breaks in the trend lines are missing samples, not idle time.

This section also holds the Guest agent card: hostname, operating system, kernel, uptime, the addresses configured inside the guest and filesystem usage. It needs the guest agent running inside the machine — machines the platform builds have it, machines adopted from another panel may not. Nothing else on the page depends on it. Use Read the guest to refresh it.

Graphs

CPU usage, memory in use, network throughput, disk throughput, disk operations per second, and monthly transfer, over a history window you choose. Figures come from the hypervisor's own counters, sampled by the agent. Network rates are in bits per second; disk throughput is in bytes per second. A machine created moments ago, or one on a node that has not reported recently, has nothing to draw yet.

Network

Every interface with its bridge, its hardware address and the ceilings actually applied to it (which may differ from what the plan advertises), plus the reverse-DNS entry. Below that, Allocated addresses lists every address reserved for this machine in IPAM.

Manage IPs allocates more or releases them:

• Allocate — choose how many and their purpose. The address is reserved immediately; configuring it inside the guest is a separate step unless the machine was built with first-boot configuration.
• Release this address — one address goes back into the pool and can be handed to another customer straight away. The guest is not reconfigured, so it keeps using the address until somebody changes it inside the machine.
• Release all IPs — every address at once, behind a confirmation checkbox. Networking breaks until new addresses are assigned.

Storage

Every disk with its target, storage type and pool, and its per-disk input/output ceilings. Resize grows a disk. A disk can be grown but never shrunk — shrinking a filesystem the guest still has mounted destroys it, and the panel always refuses.

Snapshots live here too.

• Take a snapshot — give it a name (letters, digits, dot, dash or underscore, starting with a letter or digit) and a note about why. Include RAM captures memory as well, so reverting brings the machine back still running from that exact instant; without it the machine comes back as if it had been powered off. Only a running machine has memory to capture.
• Revert — the machine goes back to how it was when the snapshot was taken. Everything written since then is discarded — files, database rows, mail, logs — and there is no way to get it back.
• Delete — the machine is untouched; only the ability to go back to that point is lost, and the disk space it was holding is freed.
• Sync — re-reads the snapshot list from the hypervisor. A snapshot removed directly on the node stays listed until you sync.

The number of snapshots one machine may hold comes from its plan, or from the module default when the plan sets none. When every slot is used, the panel tells you to delete one first. Snapshots discovered on the host that the panel did not create are marked external and must be handled on the hypervisor.

Warning: a snapshot is not a backup. It lives on the same node's storage as the machine it protects, and it holds pool space for as long as it exists. If the node is lost, so are its snapshots.

Console

The machine's screen, opened in a floating window that keeps running while you move around the panel — close it with its own control, because navigating away does not end the session.

Two controls sit alongside it, because they are what you reach for when a machine will not come up:

• Boot order — the hypervisor tries each device in turn and boots from the first one that can. Reorder, add or remove Disk, CD-ROM and Network. A machine left with the CD-ROM ahead of the disk reinstalls itself on every reboot. Where nothing is recorded, the machine boots in whatever order its own definition specifies.
• ISO / optical drive — attach an image that already exists on the hypervisor, by absolute path ending in .iso or .img. Nothing is uploaded from the panel and the guest's disks are not touched. Eject empties the drive.
• Rescue mode — puts the CD-ROM ahead of the disk and attaches a rescue image, so the machine boots from that image instead of its own operating system. The disk is never written to, resized or re-imaged — it is simply not booted from, which is what makes it safe to mount and repair from inside the rescue system. Turning rescue off restores the boot order recorded when it was turned on and ejects the medium. The machine has to be restarted for either change to take effect.

Activity

Every instruction the panel has sent to an agent about this machine, newest first, with what the agent answered. Old entries age out of the retention window set in VPS Settings.

Settings

• Identity — rename the machine (panel-side only; the name it has on the hypervisor is unchanged and stays visible on the Overview section), tags, and whether it should start automatically after a node reboot. That last one records your intent for the next operator; the panel does not write it to the hypervisor.
• Notes — staff-only, never shown to the customer.
• Billing service — which service this machine fulfils. Link to service searches your services by name, product and customer e-mail; services that are already fulfilled or in a status that cannot be linked are refused. Unlink is bookkeeping only — the machine keeps running, the service simply loses its machine. A machine with no service is not being charged for and does not appear on any customer's panel.
• CPU ceiling and I/O ceilings — applied to the running machine. Blank or zero means no ceiling. These do not change the plan; where the plan advertises a value, the panel shows it beside the field.
• Suspension — flag the machine as suspended with a reason, recorded for staff only and never shown to the customer. Powering the machine off is a separate tick, so a machine can be flagged without interrupting it yet.
• Irreversible actions — migrate, reinstall, reset password and delete, each with a one-line description of what it does.

Actions

Power

Start, Shutdown, Reboot, Pause, Resume and Power off (hard). Each is confirmed, and the confirmation says plainly what happens: shutdown asks the guest to stop cleanly, reboot interrupts service for as long as the boot takes, pause freezes the machine in place, and a hard power off cuts the power without telling the guest — the same as pulling the cord, so filesystems can be left dirty. The panel steers you towards Shutdown unless the guest has already stopped responding. You may set a Graceful window in seconds for this action, or leave it blank to use the module setting.

Reinstall OS

Overwrites the operating-system disk with a fresh image. Everything on it is gone, and you must type the machine's name to confirm. Choose the image (only the machine's plan catalogue is offered; anything else is refused when the job is queued) and optionally a new root password applied on first boot. There is also a custom image address, which skips the plan's catalogue and has no checksum — and the customer keeps being billed for the plan they bought.

Reset password

Sets a new password inside the guest through its guest agent. Nothing is written to disk and the machine is not restarted. Leave the Guest user blank and the account is resolved from the guest's operating system. This requires the guest agent to be running — machines built by the platform have it, machines adopted from another panel may not. Windows guests enforce their own complexity rule, and a password that fails it is refused inside the guest after the reset has started.

Show root password

Reveals the stored credential for a machine linked to a service. Every reveal is recorded against your account, and the value hides itself again after a minute. The panel is explicit about the cases where it cannot help: no password was ever recorded (typical of machines adopted from another panel, which keep the password the customer already had), the stored value cannot be read, or a reset did not complete so the guest may not be using it. In each case the answer is to reset the password to store a known one.

Reconfigure

Changes cores, memory or disk size. Some changes apply immediately; where they cannot, the panel says which ones take effect at the next restart. Disks can only grow.

Migrate to another cluster

Moves a running machine to another node. The destination must be in the same group as the source, and the group's keys must be current — if there is no other active node in the group, the panel tells you to add one to the group first.

You must choose a Storage mode:

• Shared storage — the disks are visible on both nodes; nothing is copied.
• Local disks — the disk contents are streamed to the destination. Slow, and the right answer for storage that lives on the node itself.

Warning: choose carefully. Picking Shared storage for local disks fails the migration; picking Local disks for genuinely shared storage can corrupt the disk. When unsure, check how the cluster's storage is set up first.

The machine stays running throughout, and progress is visible on the job.

Delete

Two forms, and the menu makes you pick one:

• Delete VM (keep disk) — the machine is removed from the hypervisor but its disk stays on the node. Nothing is wiped.
• Delete VM and wipe storage — permanently destroys the disk. You must type the machine's name. This cannot be undone.

When an action is unavailable

The panel disables an action and says why rather than letting it fail:

• Cluster is frozen — over the licensed cluster count. Power actions, console and delete still work; nothing that adds capacity does.
• Cluster is unreachable or This cluster has no provisioning agent — nothing can be sent to the hypervisor.
• This VM has no domain yet — it is still being built.
• No power action applies while this VM is ... — the action does not exist in the machine's current state.
• This VPS has been terminated — it is waiting for its disk wipe.
• On hardware another panel owns, power actions work; console, reinstall, password reset and snapshots are only available if that panel has granted full management.

What the customer can do

On their service page in the client portal, a customer with a VPS sees the machine's specification, every address allocated to it, its operating system, and live usage. What else appears depends entirely on the switches in VPS Settings:

• Power buttons — Start, Stop and Restart, each drawn only if you allowed that verb. With nothing allowed, the customer gets no power buttons at all.
• Console — the machine's screen, opened from a card on the service page.
• Reinstall — the customer picks an operating system from the plan's catalogue and a new root password, and confirms. The page then polls until the build finishes, which for a large image can legitimately take hours.
• Root password reset.
• Snapshots — off unless you switch it on, because snapshots consume node storage nobody is billed for.

If the machine is on hardware another panel owns, or on a frozen cluster, the customer sees a plain notice that management is limited and to open a ticket — power controls keep working.

Everything else in this article is staff-only: creating and deleting machines, migration, resizing, address management, boot order, ISO and rescue media, resource ceilings, linking to a billing service, tags, notes and suspension. Customers never see staff notes or a suspension reason. See Managing Your Services for the customer's view.

Where to go next

• Agent Jobs — when an action "did not happen", the answer is there.
• Clusters — the node a machine sits on.
• Service Details — the billing side of a sold machine.