Droplets
Create and configure virtual machines on hosted or owned capacity.
Droplets are virtual machines with a configured CPU profile, memory allocation, root volume, and network interfaces. Infra schedules them onto eligible nodes and records lifecycle changes as operations.
Create a droplet
Open Compute → Droplets → Create droplet in a project. Select the region, image, size, storage policy, network, and initial access settings. The placement preview checks quotas, pool eligibility, and available capacity before submission.
Choose region and image first. Those selections constrain available sizes, firmware, CPU profiles, and minimum root disk size. A size specifies a positive integer vCPU count and integer MiB memory; it must be offered by the selected pool. CPU profiles determine where that guest can move later.
Select a root-volume size in integer GiB and an eligible storage policy. Add at least one project network in the same region. A public address is optional and defaults to none. SSH keys and user data apply through the image's initialization support, not through an implicit administrator account inside every guest.
Set labels for ownership and policy selection. Deletion protection and spot capacity both default to false. Review spot retention before using a preemptible guest for any stateful application.
The Droplet API lists exact field names, omission behavior, and accepted request structure.
Unsupported sizes are rejected rather than rounded. The available size list reflects region and pool capabilities. A custom image can impose a larger root disk or different firmware requirement.
Creation lifecycle
Infra allocates an ID, reserves quota and capacity, creates storage and network dependencies, and starts the guest. The droplet moves from Provisioning to Running when the hypervisor reports execution. Guest initialization and application readiness are separate conditions.
A provisioning failure leaves an operation with its failed stage and cleanup result. A resource requiring investigation can remain visible in Error. Inspect retained volume and address IDs before creating a replacement.
Configuration changes
Display name, labels, backup selection, and deletion protection can change while the droplet runs. CPU and memory changes use resize. Storage expansion uses the volume workflow. Network changes can interrupt guest connectivity.
A configuration revision identifies the accepted configuration. The observed state shows what the agent has applied. Conflicting lifecycle operations are rejected while another operation holds the droplet lock.
Persistence
Stopping a droplet preserves attached persistent volumes and reserved addresses. It does not cancel their usage charges. Ephemeral disks and spot workloads have different retention behavior; read Spot droplets before using them for application state.
A droplet's region cannot be edited. Live migration moves placement within its region. Recovery creates replacement resources elsewhere from copied data.
Operate a droplet
Use configuration history to compare accepted revisions and cloning to create an isolated data copy. Use Guest access for SSH and serial console procedures, Lifecycle for stop, restart, resize, and deletion, and Metrics to distinguish host observations from guest measurements.
Hosted compute prices list supported size classes and specialized hardware. Root storage is billed separately from running compute; review both meters in the creation form.