Servers, pricing, and lifecycle
Create an isolated Autodisc Server, understand its quote, and operate it safely.
An Autodisc Server is optional isolated capacity for workloads that need a stable CPU/RAM shape, host-level Docker tools, SSH, custom networking, or Docker Swarm. Ordinary managed projects do not require a Server.
Launch status, checked 2026-07-13: the dashboard, signed quote, canonical ledger, and guarded lifecycle API are implemented. Real provider provisioning remains disabled until Autodisc enables a production provider. A disabled provider fails closed; it does not create an unmanaged machine.
Create a Server
- Open Servers at
/dashboard/serversand choose Create Server. - Enter a unique name and select an available region and shape.
- Choose persistent root storage and whether to reserve public IPv4. IPv6 is included when it is available in the selected region.
- Choose the quota action. The safe default stops the Server when spendable Server credit reaches zero.
- Review the signed quote, accept it, and create the Server before the displayed expiry time.
- After the Server is running and has completed enrollment, attach account SSH keys and configure Docker or Swarm from the Server-scoped pages.
The browser cannot select a physical host, submit a provider token, attach a remote IP, or create a provider resource directly. Creation is account-scoped, permission-checked, idempotent, and audited.
Pricing and the two ledgers
Server charges and managed-service GB-hours are separate:
- applications, Compose services, and managed databases use the account's GB-hour bank;
- Server CPU, RAM, storage, reserved IPv4, and backups use Server compute credit or explicitly enabled overage;
- Server RAM is never also deducted from GB-hours.
Current launch entitlements, verified 2026-07-13:
| Plan | Included Server compute credit per billing period |
|---|---|
| Free | $0 |
| Builder | $15 |
| Small | $25 |
| Premium | $75 |
The launch-v1 code defaults, verified 2026-07-13, price persistent storage at
$0.04 per GB per 730-hour month and a reserved public IPv4 at $1.00 per
730-hour month. CPU and RAM rates come from the selected shape catalog. These
are dated defaults, not a promise for a future quote: the signed quote shown in
the dashboard is authoritative and includes the exact per-second, hourly,
730-hour, and stopped rates. Its default lifetime is five minutes.
Charges accrue at per-second precision. CPU and RAM accrue while the Server is running. Storage and a reserved IPv4 continue to accrue while it is stopped. Credits do not roll over and have no cash value.
States
The dashboard presents one canonical state:
| State | Meaning |
|---|---|
draft or quoted | Configuration or quote exists; no running Server yet. |
provisioning, bootstrapping, validating | Capacity is being created, enrolled, and checked. |
running | The Server is available and CPU/RAM charges accrue. |
stopping, stopped, starting, restarting, resizing | A requested lifecycle operation is in progress or complete. |
quota_hold | Spendable Server credit is exhausted or a configured cap prevents start. |
degraded or failed | Health checks or provisioning did not complete. Review the operation detail before retrying. |
deleting, deleted | Selected runtime resources are being released or have been released. |
Provider-specific diagnostic states are internal and do not replace these customer states.
Stop, start, and delete
Stop gracefully shuts down the runtime and releases billable CPU/RAM. It preserves the Server identity and persistent storage. Storage, retained backups, and a reserved IPv4 can still cost money.
Start checks remaining credit and spend policy before reserving capacity and
booting the Server. A Server on quota_hold cannot start until credit resets,
eligible overage is enabled, or the blocking cap changes.
Delete is destructive. Review the affected workloads, volumes, addresses, snapshots, and backups. Runtime CPU/RAM stops accruing; storage and IPv4 stop only after those resources are deleted or released. Backups retained separately continue under their own retention and billing until explicitly deleted.
Quota exhaustion never deletes persistent data by itself. It stops eligible running Servers when overage is off, keeps export/delete/recovery controls available, and continues to show retained-resource cost.
Continue with Server networking, SSH, Docker, and Swarm and Server backups and recovery.