An upgrade gives a server a larger size: more vCPUs, memory or disk. It is upgrade only: nothing can be lowered, now or later, because a disk cannot shrink. Administrators upgrade any server and resellers the servers of their customers. Customers cannot upgrade; a billing module upgrades through the API (POST /servers/{id}/upgrade) with an administrator's or reseller's key.
The node needs an agent with vm.resize (0.15.0 or later); from 0.16.0 it also reports disk sizes, which settles unclear upgrades without sending them again.
Upgrade a server#
- Open the server and go to the Manage tab.
- On the Upgrade row, click Upgrade. The upgrade page opens.
- Choose one of the two ways:
- Plans: the plans offered for the server's owner in its location. Click Choose on one. A plan that would lower anything, or is the same size, is listed disabled with the reason ("Disk 20 GB is smaller than the server's 40 GB."). The server moves to the plan: its sizes, traffic allowance, what happens past it, and speed limits.
- Or sizes of its own: type vCPUs, Memory (GB) (steps of 0.5) and Disk (GB), each filled in with the current value, then Review these sizes. The server keeps its plan for traffic and speed limits; the sizes become its own.
- Check What changes ("2 → 4 vCPUs, 2 → 4 GB memory, 40 → 80 GB disk").
- Type the server's name under Confirm and click Upgrade server.
The page says "Upgrading…" and the server's controls wait until the node answers.
What happens on the node#
- vCPUs or memory change: a running server is shut down (powered off after 60 seconds if it does not stop), its disk grown, its definition updated, and it is started again. This takes about a minute. A stopped server is changed and left stopped.
- Only the disk grows: on a running server the node grows the disk live, without a restart, where it can (not on a SATA disk). Otherwise the server is shut down and started as above.
- Every other setting of the server stays: drives, ISOs, boot order, network card, firewall, speed limits.
- If the agent stops in the middle, it finishes the resize at its next start and starts the server if it was running.
Inside the server#
- Cloud images grow their partition and file system at the next boot (cloud-init
growpartandresize_rootfs). - After a live disk grow, the guest sees the larger disk at once. Grow the file system now, or wait for the next restart:
growpart /dev/vda 1
resize2fs /dev/vda1 # ext4
xfs_growfs / # XFS instead
- Windows: Disk Management › Extend Volume.
- Installed from an ISO: the server gets the larger disk, but its partitions and file system must be grown inside it by hand.
Extra disks are grown on the server's Storage tab; see Disks, CPU and GPU.
What is checked first#
Before anything is sent the panel refuses, in plain words:
- a suspended server, one in rescue, being built or deleted, in error, with a task under way, or with an upgrade not confirmed yet;
- a server whose node does not answer;
- a node without room for the increase under its overselling settings (only the increase counts; a reseller reads "does not have room", not the node's figures). An administrator cannot oversell a node with an upgrade: raise the node's overselling settings instead;
- a reseller pool without room for the increase in vCPUs, memory or disk;
- a disk increase larger than the free space of the server's storage.
Upgrades sent and not confirmed yet count as used, on the node and in the pool.
When the node does not confirm#
If the task runs out of time or the agent loses it, the upgrade stays pending and still holds its increase. The page says "Upgrading… waiting for the node to confirm the new size." Each node poll checks the server's sizes and takes the upgrade once the node lists them.
After 6 hours nothing more is sent and administrators see Needs attention on the server's page. Check the server on the node, then click Forget this upgrade: the server keeps the sizes shown and the held increase is freed. A larger disk the node lists later is taken anyway.
Suspend a server#
Suspending stops a customer from using a server, for example for an unpaid invoice, without deleting anything. Administrators suspend any server; resellers the servers of their customers. Users cannot suspend or lift.
- Open the server and go to the Manage tab.
- On the Suspend row, click Suspend.
- Type a Reason (required, up to 200 characters). The owner reads it.
- Click Suspend server.
What happens:
- A running server is paused on the node and keeps its memory; a stopped one stays stopped.
- The status becomes Suspended (amber). The owner reads "Suspended:
. Contact your provider." Power, console, ISO drives, settings, reinstall, rescue, password reset, network and address changes are closed; only Delete server stays, for whoever may delete it. An open console window is closed. - The node refuses every task but delete, keeps the server paused across an agent restart, and starts it paused after a node reboot.
- Queued tasks of the server are cancelled; a running one finishes first, then the node is told.
- The server's memory still counts as sold, and the server still counts against its owner's and reseller's limits.
To change the reason, click Change reason. An administrator who changes a reseller's suspension takes it over.
Lift a suspension#
On the Manage tab, click Lift suspension. The node resumes a paused server where it was.
- A reseller lifts only a suspension it placed itself ("An administrator suspended this server. Only an administrator can lift it.").
- A server suspended for its traffic allowance shows Lift traffic suspension to administrators; otherwise it lifts itself when the month resets. A server can hold both suspensions at once; each button lifts one, and the node resumes the server only when neither is left.
If the node refused the pause, staff read "Not paused on the node yet:
Audit: vm.suspended and vm.unsuspended. Billing modules suspend and unsuspend through the API (POST /servers/{id}/suspend and /unsuspend); see Billing modules.