Managing your server

Everything here lives under VServer in the portal, split across the tabs Overview, Console, Network, Snapshots, Metrics and Settings. If you have several servers, the row of names above the tabs switches between them.

Power

Overview → Power controls has Start, Restart and Stop.

Stop asks the operating system to shut down cleanly. If the guest has not powered off within two minutes, it is cut instead, so a hung machine still stops. Restart is a reboot, not a stop followed by a start.

A stopped server still costs money

Stopping is not the same as cancelling. The plan is billed for a stopped server exactly as for a running one, because the memory and disk stay reserved for it. See Billing.

The portal refreshes the live state every 45 seconds, and a background check reconciles what the portal shows with the real state once a minute. So a status that lags a power action by a few seconds is normal.

The browser console

Console → Connect attaches a keyboard and a screen to the machine. It runs below the network stack inside your server, which is what makes it the right tool when SSH is what is broken: a wrong firewall rule, a mistyped sshd_config, a boot that stops at the initramfs prompt.

It works while the server is Active or Stopped. There are buttons for Ctrl+Alt+Del, for pasting the clipboard into the session, and for fullscreen.

Each session is opened with a one-shot ticket that is valid for 60 seconds, so a Connect you clicked and left sitting will not work later. Click it again.

Metrics

Metrics shows CPU, memory, disk I/O and network for the last hour, day, week or month. It is measured from outside the server, so it keeps working when the machine is unresponsive, and it does not know anything about processes.

Email alerts

Two alerts are on by default and are switched under Account settings → Notifications.

• Server down. Sent when your server is found stopped although you had it running. Stopping it yourself does not trigger it. The check runs once a minute; you get at most one email per server per 6 hours, re-armed when the server comes back.
• Disk almost full. Sent when any real filesystem on the server reaches 90 % used. The check runs every 15 minutes; at most one email per server per 24 hours, re-armed once the filesystem drops back below the threshold.

The disk alert reads the filesystems through the guest agent. No agent means no disk alert, silently, which is one more reason to leave it installed. See The guest agent.

Renaming a server

Settings → Hostname changes the name the portal, the alert emails and the server list use.

Renaming does not rename the operating system

The hostname inside a running server is set once, at first boot. Changing the name in the portal does not reach into the machine and change it. Do that yourself if you care:

hostnamectl set-hostname new-name

Snapshots

Snapshots takes a point-in-time copy of the disk that you can return to. You can keep 3 per server. A fourth fails with a generic error in the portal; if creating a snapshot fails and nothing else is wrong, that is almost always why.

• A snapshot captures the disk only. Memory is not included, so returning to one is like a power cut at that instant, not a resume.
• Restore to this point discards everything written since the snapshot was taken, with no way back. Check the power state on Overview afterwards.
• Snapshots taken automatically before a plan change are labelled Automatic and do not count against your 3.

A snapshot is not a backup

Snapshots sit on the same storage as the server itself. They protect you against a bad upgrade, a wrong rm or a broken configuration change. They do not protect you against losing the machine, and a reinstall or a deletion destroys them along with it.

Backups

Separately from your snapshots, whole-server backups are taken off-site by the platform. They exist for the case the snapshots cannot cover.

There is no restore button. Restoring from a backup is a support request, so plan around snapshots for anything you expect to need back quickly, and keep your own copies of data you cannot lose.

Reinstalling

Settings → Reinstall OS rebuilds the server from a fresh image. You type the hostname to confirm, because everything on the disk goes.

You pick the image, and optionally paste one or more public keys, one per line. These are pasted keys, not your stored ones: the field is a text box. Leave it empty and the new system comes up with no key at all, in the same state as an order placed without one.

What survives: the IPv4 address, the plan and the disk size.

What does not: the contents of the disk, and every snapshot on the server.

A reinstall gives you a new IPv6 block

The IPv4 address is kept. The IPv6 /112 is derived from the rebuilt machine and comes out different, so anything pinned to the old IPv6 address, AAAA records included, has to be updated afterwards.

Changing the plan

Settings → Change plan moves the server to another plan. The server is shut down, reconfigured and started again, so it is offline for the duration, and only one change can run at a time.

A safety snapshot is taken first, unless you tick Skip safety snapshot. It is removed again when the change succeeds and kept when it fails, so there is something to roll back to.

The disk can only grow, and growing is permanent

Grow disk to new plan size is ticked by default. Once the disk has grown it can never be shrunk, so from that point no smaller plan is reachable.

Untick it if you might want to go back down later: the server then runs the new plan's CPU and memory on its current, smaller disk. Attempting a shrink is refused outright:

disk can only grow (current 60G, target 40G)

After a disk grow, check that the filesystem followed the partition:

df -h /
lsblk

Most images expand the root filesystem on the next boot on their own. If yours did not, grow the partition and then the filesystem with growpart and resize2fs (or xfs_growfs), using the device names lsblk printed.

Billing switches to the new plan's rate from the moment the change lands, not at the end of the month. See Billing.

Deleting a server

Settings → Delete VServer, confirmed by typing the hostname, cancels the subscription and tears the server down.

This is immediate and irreversible. The disk, all snapshots and the IP allocation go with it, and there is no recycle bin and no grace period. Billing stops at that point, with the final part-hour on your next invoice.

What's next

• Networking for addresses and what can reach them
• Troubleshooting when something is not working