Managing Instances

Create, monitor, and control your GPU instances — start, stop, delete, view logs, check SLA, and manage SSH keys.

Overview

All GPU deployment modes (Marketplace, Dex-Cloud, Burst) create instances. Once created, manage them through a unified interface.

List Instances

from gpuniq import GPUniq
client = GPUniq(api_key="gpuniq_your_key")

# Active instances
instances = client.instances.list(page=1, page_size=20)
for inst in instances["instances"]:
    print(f"Task {inst['task_id']}: {inst['status']} — {inst['gpu_model']}")

# Archived (deleted) instances
archived = client.instances.list_archived()

Instance Details

Get full information about a specific instance:

details = client.instances.get(task_id=456)

Returns SSH connection details, hardware specs, billing info, and status.

Instance Lifecycle

Start

Resume a stopped instance:

client.instances.start(task_id=456)

Stop

Pause billing while preserving instance state:

client.instances.stop(task_id=456)

Delete

Permanently terminate the rental:

client.instances.delete(task_id=456)

Deleting an instance is irreversible. All data on the instance will be lost. Use Volumes for persistent data.

Rename

client.instances.rename(task_id=456, name="bert-finetuning-v2")

Connecting via SSH

After creating an instance, SSH credentials are available in the dashboard:

ssh root@<host> -p <port>
# Password shown in dashboard

SSH Key Authentication

For passwordless access, add SSH keys before creating the instance:

# Add key in settings
key = client.settings.create_ssh_key(
    key_name="my-laptop",
    public_key="ssh-ed25519 AAAA... user@laptop",
)

# Pass when creating the order
order = client.marketplace.create_order(
    agent_id=123,
    pricing_type="hour",
    ssh_key_ids=[key["id"]],
)

Manage Instance SSH Keys

# List keys attached to instance
keys = client.instances.ssh_keys(task_id=456)

# Attach / detach
client.instances.attach_ssh_key(task_id=456, ssh_key_id=1)
client.instances.detach_ssh_key(task_id=456, key_id=1)

Container Logs

View Docker container logs for your instance:

logs = client.instances.logs(task_id=456)

SLA Monitoring

Check uptime and availability data:

sla = client.instances.sla(task_id=456)

Returns 24-hour and 7-day uptime segments and percentages.

Pending Jobs

Track deployment jobs that haven't completed yet:

jobs = client.instances.list_pending_jobs()

# Cancel a pending job
client.instances.cancel_pending_job(job_id="abc-123")

Best Practices

Stop instances when not in use — billing pauses immediately
Use longer pricing types for production workloads (20-40% savings)
Attach volumes for persistent data instead of relying on instance storage
Monitor SLA for critical workloads to track uptime
Use verified providers for maximum reliability

Was this page helpful?

Last updated Feb 22, 2026

Built with Documentation.AI