VPS Settings

VPS Settings holds the module-wide behaviour: what a customer may do to their own virtual server, how a node is chosen for a new one, what counts as a problem worth flagging, and how long history is kept. Every setting on this page applies to the whole module. Anything that belongs to one machine, one node or one group is set there instead.

Opening the page

Open VPS in the admin sidebar, then the Settings tab.

How the page works

• Settings are grouped into sections, each with a short note explaining what the group actually affects.
• A search box matches every setting in the module, and a Changed only filter narrows the page to what you have edited but not yet saved.
• Each row shows its current value, its default and, for numbers, the allowed range. Reset to default puts one row back.
• Edits are held until you save. A banner counts unsaved changes; the save button says how many it will write; Discard changes abandons them.
• The save is all or nothing. A value that is not valid — a blank number, or malformed content in a structured field — blocks the save until you fix it, and the row says so. A number outside its range is clamped, and the row tells you the value it will actually save.
• Saved settings take effect immediately across the whole module.

Note: if the page says this build exposes no VPS settings, the module has not finished being set up on your panel yet.

What customers may do

These decide what the client portal offers somebody on their own virtual server. An administrator is never restricted by them. Turning reinstall off is how a managed-hosting provider stops a customer wiping a server the provider supports.

Customer-facing permissions

Setting	What it does
Power buttons	Which power actions a customer may perform on their own machine. The choices are start, shutdown and reboot; all three are allowed by default. Tick nothing and customers get no power buttons at all — a valid choice, and the panel confirms that is what it will save.
VNC console	Whether a customer may open the screen of their own machine. On by default.
Reinstall	Whether a customer may reinstall their own machine. On by default. A reinstall destroys everything on the machine, so off is the right answer for managed hosting.
Root password reset	Whether a customer may reset the root password of their own machine. On by default.
Snapshots	Whether a customer may take and restore their own snapshots. Off by default, because snapshots consume node storage nobody is billed for. The per-machine ceiling is under Limits below, and a plan can set its own.

Shared hardware

The section is headed Shared fleet on screen. It applies to hypervisors another FluxBilling panel owns and shares with you.

Setting for hardware another panel owns

Setting

What it does

Shared auto claim

Off by default. When on, this panel automatically takes over a machine on shared hardware when it has a service recorded against the source panel's own identifier for that machine — an exact identity match, never an address, because addresses are recycled between customers. A claimed machine is billed here and gains power, console and reinstall from this panel. Leaving this off and claiming machines deliberately is the safer route.

Placement and sell ceilings

These are the last fallback, used only where nothing more specific applies.

Placement defaults

Setting	What it does
Server selection	How a node is chosen for a new machine when the product is not pinned to one. The choices are None (first available), Most available RAM (the default), Most available disk, Most available cores, Most available IPv4, Least number of VPS and Least system load.
Fallback sell ceiling	Default multipliers for memory, vCPU and disk, used only by a node that sets neither a multiple nor a fixed amount of its own. Empty means no limit. Most operators leave this blank and set the ceiling per node, where the panel can show what it works out to in real units.
Freeze order	The order clusters are frozen in when the licensed cluster count is exceeded — the first listed freezes first. Unset means newest first, and any cluster you do not list freezes last. Clusters shared with you are never listed: their owner's licence decides when they freeze. Freezing is a licence guard, not an outage — a frozen cluster keeps every machine running and manageable, and only creating, reconfiguring and adopting are refused.

A panel on this section shows the inheritance chain in full, labelled Cluster → Group → Fleet default: a new machine resolves its node selection and its ceilings from the most specific setting that has one — the cluster first, then the group it belongs to, then the values above, which is the step the panel calls Fleet default. It also counts how many clusters and groups currently override each one.

Note: a sell ceiling is enforced against one machine's hardware, so it belongs on the node — Clusters → the node → Settings. See Clusters.

Lifecycle

What the platform does to a virtual server on its way from suspension to termination.

Lifecycle settings

Setting	What it does
Shutdown window before power off	How long the guest is given to shut itself down cleanly before a suspended machine is powered off hard. Default 300 seconds; allowed 0 to 86400.
Default DNS resolvers	The resolvers handed to a new machine when the subnet it is built on defines none of its own. Comma separated, up to eight addresses. Leaving it blank falls back to the built-in pair.

Warning: this is not the billing grace period. Billing settings decide when an unpaid service is suspended or terminated. The setting above decides only what happens to the guest once that decision is made. Termination destroys the machine and wipes its disk — immediately, with no retention window and no way back — and it runs unattended: anything left suspended past the termination grace period is terminated automatically, as is a scheduled cancellation and an hourly machine that runs out of credit.

Alert thresholds

These define what "needs attention" means on the module's Overview tab. Nothing here changes what a node does — only when the panel calls it out, so setting them too loosely hides a problem rather than fixing it.

Alert thresholds

Setting	What it does
SSD life remaining	Warn when a solid-state disk reports this much rated write life left or less. Default 10%; allowed 0 to 100.
Memory oversell ratio	Warn when a node commits more than this multiple of its physical memory. Default 1.5; allowed 0.5 to 20.
Host memory used	Warn when host memory usage reaches this percentage. Default 90%; allowed 1 to 100.
Storage pool full	Warn when a storage pool reaches this percentage full. Default 90%; allowed 1 to 100.
Node stops reporting	Treat a node as not reporting after this many minutes without a sample. Default 30 minutes; allowed 1 to 1440. Agents sample every 5 minutes, so anything below about 10 will produce noise.

History retention

History costs disk. Every extra day of samples, traffic totals and job records is growth on your platform database.

Retention windows

Setting	What it does
Telemetry samples	Days of per-machine and per-host samples to keep — the data behind every graph. Default 30 days; allowed 1 to 3650.
Traffic totals	Months of daily traffic totals to keep. Default 13 months; allowed 1 to 120.
Agent job history	Days a completed or failed job is kept. Default 3 days; allowed 1 to 365. See Agent Jobs.

Note: each of these has a floor above zero, so the purge can never be switched off entirely.

Timeouts

How long a request waits on an agent or on another panel before it answers "pending" and lets the work finish in the background. Longer is not safer — this is a worker held open.

Timeouts

Setting	What it does
Power action wait	How long a power action waits for the agent before returning as pending. Default 30000 ms; allowed 1000 to 300000.
Resize wait	How long a resize waits for the agent before returning as pending. Default 60000 ms; allowed 1000 to 600000.
Shared-cluster peer timeout	Timeout for calls to the panel that owns shared hardware. Default 30000 ms; allowed 1000 to 300000.
Takeover nodes at once	How many source machines the import wizard works on at the same time. Default 2; allowed 1 to 16. See VPS Panel Migration.

Tip: a request timing out here does not cancel the work. The job keeps running and the screen catches up when it finishes; these values only decide how long the panel holds the browser waiting.

Limits

Limits

Setting	What it does
IPs per allocation call	The most addresses one allocation may attach to a machine. Default 256; allowed 1 to 4096.
Snapshots per VPS	The snapshot limit for a machine whose plan does not set one. Default 5; allowed 0 to 100. A plan that carries its own allowance overrides this.
Require marked ips	Off by default. When on, every virtual-server address must come from address space marked Virtual Server in IPAM, company-wide: a location with no marked space refuses to build rather than borrowing an address a physical server may also be handed. With it off, a node from which marked space is reachable still allocates only from marked space, while a node with none keeps using the general pool — so nothing changes for a company that marks nothing.

Common tasks

Make VPS management read-only for customers

1. Untick every entry under Power buttons.
2. Turn off VNC console, Reinstall and Root password reset.
3. Leave Snapshots off.
4. Save.

The client portal then shows the machine and its usage but offers no controls. Administrators are unaffected.

Keep more graph history

Raise Telemetry samples. Every extra day is measurable growth on your database, so raise it deliberately and watch your disk.

Decide which clusters freeze first

Open Freeze order and add clusters in the order they should freeze, moving them up and down. Put your least valuable capacity first. Anything you do not list freezes last, newest first.

Where to go next

• VPS Overview — how the module fits together.
• Clusters — per-node settings, including sell ceilings and storage pools.
• Groups — per-group placement.
• VPS Plans — per-plan snapshot allowances and resource ceilings.