Guide · For administrators
Plans and operating systems
Plans
In Clients & plans › Plans you decide what you sell. Changing a plan does not touch existing VPSs: to move them to the new resources use Change plan on the VPS.
| Field | What it does |
|---|---|
| vCPU, RAM, disk | VPS resources |
| Network speed | interface limit in Mbit/s (0 = unlimited) |
| Monthly traffic | GB included per month, in plus out (0 = unlimited) |
| When exceeded | slow down until next month (to the reduced speed), suspend until next month, or notify only |
| Snapshots | how many snapshots the customer can keep (0 = none) |
| Backups kept | how many backups are kept (0 = no backups for the customer) |
| Automatic backups | no, every night, once a week; and where to store them |
| IPv6 | one IPv6 address per VPS, on nodes that have an IPv6 network |
| Fast writes | writes that programs confirm (databases) don't wait for the disk: 8–9 times faster. If the node loses power suddenly the VPS loses the last few seconds of data. Off by default; ZFS nodes only. It can also be turned on for a whole node (see Nodes) |
| High availability | the plan's VPSs are copied every few minutes to a reserve node and, if their node suddenly goes down, they restart there with the same IP. Uses space on the reserve node; ZFS nodes only. See High availability |
| Disk and CPU limits | disk read and write (MB/s and operations per second), maximum CPU use and priority: see below (0 = no limit) |
Traffic resets on the first day of the month: VPSs slowed down or suspended for traffic go back to normal on their own. At 90% the customer gets a notice.
Traffic is measured on the node, on the VPS network cards (public, IPv6, extra IPs and private networks), every 30 seconds: what happens inside the VPS (agent stopped or changed, Docker, VPNs) does not change the count. A panel restart or update does not lose traffic.
Disk and CPU limits
With the CPU sold up to 4 times the server's threads and shared disks, a VPS that works non-stop can slow down every other VPS on the same node (the "noisy neighbour"). Since 1.14 every plan can have brakes, in the Disk and CPU limits box:
| Field | What it does |
|---|---|
| Disk read | MB/s and operations per second (IOPS) for reads (0 = no limit) |
| Disk write | MB/s and operations per second for writes (0 = no limit) |
| CPU: maximum sustained use | the share of each vCPU the VPS can use, from 10 to 100% (0 = no limit). With 2 vCPUs and 50% the VPS uses at most one whole CPU, even when the server is idle |
| CPU priority | normal, low (when the node is busy it gives way to the others: it counts a quarter) or high (it counts double). With an idle node nothing changes |
For reads and for writes you can set MB/s, operations per second or both: whichever is reached first applies. For a database operations per second matter; for backups and file copies MB/s. 1 MB/s = 1,000,000 bytes per second.
Disk limits apply to every disk of the VPS separately: the main one and the extra ones, attached volumes too. With 200 MB/s for reads and two disks, each disk reads at most 200 MB/s. They work with ZFS and LVM; they do not apply to disks on dir pools, which are files and not block devices.
The panel applies the limits on the node, over SSH (see Nodes): every VPS of the node goes into its own group (cgroup) and the limits apply within 30 seconds, even while the VPS is running. When you save the plan all its VPSs get them within seconds. After a start the VPS runs without limits for 30 seconds at most. Without SSH access to the node the limits do not apply. They can be turned off for everyone in Settings › Nodes and overbooking › VPS groups on the nodes.
Unlike CPU, RAM and disk, the limits also apply to the plan's existing VPSs. The custom resources of a VPS can have limits of their own (see Customers and VPSs); a plan change puts back those of the new plan. Provider VPSs (Hetzner, Contabo…) do not have these limits.
On the node the plan's RAM and disk count as assigned, even if the VPS uses less (vCPUs are not counted). How much extra to sell compared with the node's real resources is set by each node's overbooking.
Operating systems
Systems have their own page, Systems: KoaVM catalogue, your own templates, Windows from ISO, the ISO library and apps. There you also choose which plans can use each system and the minimum resources. See Systems, templates, ISOs and apps.
Every Linux VPS starts with cloud-init: hostname, static network (IPv6 too), root password, SSH key if given, SSH server running. Windows VPSs get the Administrator password, network and remote desktop from cloudbase-init.