Cloud Instance Management

Overview

Once provisioned with skytells cloud deploy, cloud instances are managed with the commands documented here: listing, inspecting, controlling power state, and destroying.

skytells cloud ls

Lists all cloud instances in the linked project.

Syntax

skytells cloud ls [--json]

Example

skytells cloud ls

┌─────────────┬────────────────┬────────┬─────────┬──────────────┐
│ ID          │ Name           │ GPU    │ Status  │ Region       │
├─────────────┼────────────────┼────────┼─────────┼──────────────┤
│ inst_abc001 │ llm-trainer    │ H100   │ running │ us-east-1    │
│ inst_abc002 │ inference-api  │ A100   │ halted  │ us-west-2    │
│ inst_abc003 │ api-worker     │ —      │ running │ us-east-1    │
└─────────────┴────────────────┴────────┴─────────┴──────────────┘

# Find running GPU instances
skytells cloud ls --json | jq -r '.[] | select(.status == "running" and .gpu != null) | .id'

skytells cloud inspect

Returns full details of a single instance: hardware specs, network info, status, and region.

Syntax

skytells cloud inspect <id> [--json]

Example

skytells cloud inspect inst_abc001

# Poll until running
until [[ "$(skytells cloud inspect inst_abc001 --json | jq -r '.status')" == "running" ]]; do
  echo "Waiting..."
  sleep 15
done
echo "Instance is ready"

skytells cloud start

Starts a halted instance. The instance resumes with its previous disk state.

Syntax

skytells cloud start <id> [--json]

skytells cloud start inst_abc002

skytells cloud halt

Halts a running instance. The process is stopped; disk state is preserved. The instance can be restarted with skytells cloud start.

Syntax

skytells cloud halt <id> [-f] [--json]

Flag	Description
`-f`	Force halt without draining active connections.

skytells cloud halt inst_abc001

Halting stops compute charges but the GPU reservation may still incur a standby fee. To release all resources, use skytells cloud destroy.

skytells cloud reboot

Reboots a running instance. Equivalent to halt followed by start. Use to apply low-level system configuration changes that require a restart.

Syntax

skytells cloud reboot <id> [--json]

skytells cloud reboot inst_abc001

skytells cloud destroy

Permanently destroys an instance and releases all associated resources. Disk data is irrecoverably deleted.

Syntax

skytells cloud destroy <id> [-f] [--json]

Flag	Description
`-f`	Skip the confirmation prompt.

# Destroy with confirmation
skytells cloud destroy inst_abc001

# Destroy immediately (CI/CD cleanup)
skytells cloud destroy inst_abc001 -f

Destruction is irreversible. All disk data, configuration, and compute resources are permanently removed. This cannot be undone. Ensure you have saved any work from the instance before destroying it.

Automated cost-control

Stop all halted instances at end of day:

#!/bin/bash
# Destroy all halted cloud instances
skytells cloud ls --json | \
  jq -r '.[] | select(.status == "halted") | .id' | \
  xargs -I{} skytells cloud destroy {} -f

Halt all running GPU instances:

skytells cloud ls --json | \
  jq -r '.[] | select(.status == "running" and .gpu != null) | .id' | \
  xargs -I{} skytells cloud halt {}

How is this guide?

On this page