Virtual machines
Creating a server, starting and stopping it, getting in with SSH or through the console, and deleting it.
Before you start: the project's cloud must be switched on (Getting started) and it is a good idea to have an SSH key saved.
Creating a server
In Cloud, Normal mode, Servers section, press New server. All the choices are on one screen, each with a sensible value already filled in:
- Name: the panel suggests a random one (like
brave-otter); you can change it. Lowercase letters, digits and hyphens, 2 to 40 characters, unique in the project. - Size: vCPUs and memory (for example 2 vCPU and 4 GB RAM). The available sizes are set by the zone.
- Operating system: the image. Ubuntu, if available, is already selected.
- Disk (GB): the system disk, from 10 to 2000 GB (20 if you leave it; some images require a higher minimum).
- Public IPv4 address: on, the server gets its own public address, routed straight to it; off, the server is private only (see Networking). If the project can take addresses from more than one public network, choose which in Public network. Where the public network also has IPv6, the server gets an IPv6 address too.
- SSH key: choose a saved key, or Paste a key below and paste a public key. Without a key you cannot get in: the images have no password.
- Open ports: 80 (HTTP), 443 (HTTPS), 22 (SSH). If you choose 22, you must say where it can be reached from in Who can reach these ports: an address or a network, IPv4 or IPv6 (for example
203.0.113.7/32). That source applies to all the ports chosen there. Other ports you open later, from the firewall.
Press Review the plan: you see what will be created (server, disk, address, firewall rules) and the quota afterwards. If it looks right, Create server. The first server is ready in about half a minute; meanwhile it shows Starting up.
For now, the reverse DNS name of the public address is set with the assistant, the API or the MCP, not from this screen: see Reverse DNS.
Server states
| In the panel | In the API (status) |
Meaning |
|---|---|---|
| Starting up | BUILD |
Being created |
| Running | ACTIVE |
Running |
| Stopped | SHUTOFF |
Stopped |
| Deleting | DELETING |
Being deleted |
| Error | ERROR |
Something went wrong: the reason is in the details |
The server page
Click a server to open its page:
- at the top: state, vCPUs and memory, public IPv4 and IPv6 address (with a button to copy them);
- the Start / Stop, Restart buttons and, for Can manage, Console;
- Size, Private address, Public address, Created and, if the server was created with an SSH key, the SSH line with the command to log in;
- the Firewall, with the open ports (Firewall);
- the Disks;
- at the bottom, Delete.
If one of the server's addresses is under attack, the page says so (Under attack): see Attack protection.
Starting, stopping, restarting
- Stop shuts the server down cleanly. Disk and addresses remain.
- Start starts it again.
- Restart is a clean shutdown followed by a start: for a moment the server does not respond.
A stopped server only pays for disk and public addresses; CPU and memory only count while it is running. If the account is out of credit, a stopped server cannot be started again until you top up (see Billing).
Connecting with SSH
You need three things: the server running with a public address, your SSH key added at creation, and port 22 open from your address (in the firewall, the Me button fills in the address you are connected from).
On Ubuntu images the user is ubuntu:
ssh ubuntu@<server's public address>
Other images have their own default user (for example debian for Debian). Once in, you are a user with sudo.
A server without a public address can be reached from the other servers in the same project, over the private network: for example, you log into a public server and jump from there to the private one (ssh -J ubuntu@<public> ubuntu@<private>).
Good to know. With today's images the SSH server is installed on first boot, and to download it the server must be able to get out to the internet: with its public address, or through the project's NAT gateway if there is one (see Networking). A private-only server, in a project with no way out, is created without SSH. If SSH does not answer in the first few minutes after creation, it is probably still installing.
The console
When a server's network does not respond (wrong firewall, broken network configuration), there is the console. The Console button on the server page, with two screens:
- Screen: the server's screen, in the browser. Click on the screen to type into it.
- Text: the serial console, as text.
The console:
- can only be opened by Can manage and the owner: it shows everything the server shows;
- only works while the server is running;
- stays open for at most one hour;
- every opening is recorded in the project's log.
Note: servers created here are accessed with the SSH key and have no password. The console shows the boot and the errors, but to log in from the console you need a password set inside the server (for example with sudo passwd ubuntu, while you can still get in via SSH). If you think you will need it, set it beforehand.
Disks
On the server page and in the Disks table:
- the system disk (system · deleted with the server) is deleted with the server;
- an additional disk (additional · survives the server) remains even if the server goes away.
Today disks can only be viewed: creating additional disks from the panel is not available yet.
Deleting a server
On the server page, Delete. The plan tells you what will go; to confirm you must type the server's name again, then Delete for good.
Deleting a server destroys its system disk and gives back its public address, which may go to another customer. Additional disks remain. There is no undo.
For developers
| What | Method (/api/rpc) |
MCP tool | Role |
|---|---|---|---|
| Sizes, images, public networks, Kubernetes versions | Cloud/getCatalog |
cloud_get_catalog |
viewer |
| List of servers | Cloud/getServerList |
cloud_get_server_list |
viewer |
| One server, with disks and firewall | Cloud/getServer (server_id) |
cloud_get_server |
viewer |
| Create | Cloud/createServer |
cloud_create_server |
operator |
| Start, stop, restart | Cloud/serverAction (server_id, action: start, stop, reboot) |
cloud_server_action |
operator |
| Delete | Cloud/deleteServer (server_id) |
cloud_delete_server |
operator |
| Console | Cloud/getConsole (server_id, type: vga or serial) |
cloud_get_console |
admin |
| The project's disks | Cloud/getVolumeList |
cloud_get_volume_list |
viewer |
| State of a long-running operation | Cloud/getJob (job_id) |
cloud_get_job |
viewer |
Fields of Cloud/createServer:
| Field | Required | What |
|---|---|---|
name |
yes | 2–40 characters, lowercase letters, digits, hyphens |
size |
yes | A name from Cloud/getCatalog, for example s-2-4 |
image |
yes | A name from Cloud/getCatalog, for example ubuntu-24.04 |
disk_gb |
no | 10–2000, default 20 |
public_ip |
no | true by default |
public_ip_pool |
no | An id from Cloud/getPublicIpPoolList; without it, the default network |
ssh_key |
no | A complete OpenSSH public key |
ssh_key_uuid |
no | Or the uuid of a key saved in the account (/api/rpc only, not MCP) |
open_ports |
no | TCP ports to open straight away, for example [80, 443] |
source |
for ports other than 80 and 443 | Address or network from which open_ports can be reached |
reverse_dns |
no | The reverse DNS name: see Reverse DNS. Requires public_ip |
idempotency_key |
no | A string of your own: two requests with the same key create only one server |
dry_run |
no | true: only the plan, creates nothing (Plans) |
Example, first the plan and then the creation:
curl -s https://gigakube.ch/api/rpc \
-H "Authorization: Bearer $GIGAKUBE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"Cloud/createServer","params":{"call_data":{
"name":"web-1","size":"s-2-4","image":"ubuntu-24.04","disk_gb":40,
"ssh_key":"ssh-ed25519 AAAA… me@laptop",
"open_ports":[80,443],"dry_run":true}}}'
Remove "dry_run":true (and add an idempotency_key) and the same call creates the server. The response contains a job: via MCP the platform waits up to about 25 seconds for it to finish; if it has not finished, read it again with Cloud/getJob / cloud_get_job. The server appears in Cloud/getServerList (in the items list) with status, public_ip, public_ip6, addresses, reverse_dns, reverse_dns_pending.
Cloud/getConsole responds with a url to open in the browser within 60 seconds; the session lasts at most one hour. A key that requires approval cannot open the console.