Launch your first OCI compute instance with capacity retry logic.
Use when creating a new compute instance, testing OCI connectivity, or hitting "Out of host capacity" errors on Always Free ARM shapes.
Trigger with "oraclecloud hello world", "launch oci instance", "oci compute example", "out of capacity oci".
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
Launch your first OCI compute instance with capacity retry logic.
Use when creating a new compute instance, testing OCI connectivity, or hitting "Out of host capacity" errors on Always Free ARM shapes.
Trigger with "oraclecloud hello world", "launch oci instance", "oci compute example", "out of capacity oci".
allowed-tools
Read, Write, Edit, Bash(pip:*), Bash(oci:*), Grep
version
1.7.0
license
MIT
author
Jeremy Longshore <jeremy@intentsolutions.io>
tags
["saas","oraclecloud","oci"]
compatibility
Designed for Claude Code
Oracle Cloud Hello World
Overview
Launch, list, and manage your first OCI compute instance. The most common blocker for new OCI users is the Out of host capacity error when launching Always Free ARM shapes (VM.Standard.A1.Flex). This error means the data center has no available hosts — it is not a permissions issue. The solution is a retry loop that polls until capacity becomes available.
Purpose: Get a running compute instance on OCI, including the capacity retry pattern that makes Always Free ARM shapes actually usable.
Prerequisites
Completed oraclecloud-install-auth — valid ~/.oci/config with API key authentication
Python 3.8+ with pip install oci installed
A subnet OCID in your tenancy (VCN > Subnets in the Console, or use the default VCN)
An image OCID for your region (Compute > Custom Images, or list platform images via API)
An SSH public key at ~/.ssh/id_rsa.pub (for instance access)
Instructions
Step 1: List Existing Instances
import oci
config = oci.config.from_file("~/.oci/config")
compute = oci.core.ComputeClient(config)
instances = compute.list_instances(compartment_id=config["tenancy"])
for inst in instances.data:
print(f"{inst.display_name:<30}{inst.lifecycle_state:<12}{inst.shape}")
Step 2: List Available Shapes and Images
# List shapes available in your tenancy
shapes = compute.list_shapes(compartment_id=config["tenancy"])
for s in shapes.data:
ocpus = getattr(s, "ocpus", "fixed")
print(f"{s.shape:<} OCPUs: ")
images = compute.list_images(
compartment_id=config[],
operating_system=,
sort_by=,
sort_order=,
limit=
)
img images.data:
()
35
{ocpus}
# List platform images (Oracle Linux)
"tenancy"
"Oracle Linux"
"TIMECREATED"
"DESC"
5
for
in
print
f"{img.display_name:<60}{img.id[:40]}..."
Step 3: Launch an Instance (Standard)
import os
launch_details = oci.core.models.LaunchInstanceDetails(
compartment_id=config["tenancy"],
availability_domain="Uocm:US-ASHBURN-AD-1", # Change for your region
display_name="hello-oci",
shape="VM.Standard.E4.Flex",
shape_config=oci.core.models.LaunchInstanceShapeConfigDetails(
ocpus=1, memory_in_gbs=8
),
source_details=oci.core.models.InstanceSourceViaImageDetails(
image_id="ocid1.image.oc1.iad.aaaa...", # Your image OCID
boot_volume_size_in_gbs=50
),
create_vnic_details=oci.core.models.CreateVnicDetails(
subnet_id="ocid1.subnet.oc1.iad.aaaa..."# Your subnet OCID
),
metadata={
"ssh_authorized_keys": open(
os.path.expanduser("~/.ssh/id_rsa.pub")
).read()
}
)
response = compute.launch_instance(launch_details)
instance_id = response.data.idprint(f"Launching: {instance_id}")
print(f"State: {response.data.lifecycle_state}")
Step 4: Wait for Instance to Be Running
# Poll until RUNNING (typically 30-90 seconds)
get_instance = oci.core.ComputeClient(config)
waiter = get_instance.get_instance(instance_id)
result = oci.wait_until(
get_instance,
waiter,
"lifecycle_state",
"RUNNING",
max_wait_seconds=300
)
print(f"Instance is RUNNING: {result.data.display_name}")
Step 5: Launch with Capacity Retry (Always Free ARM)
This is the pattern you need for VM.Standard.A1.Flex shapes. OCI returns Out of host capacity intermittently — retry until a host becomes available:
import time
import random
deflaunch_with_retry(compute_client, launch_details, max_retries=720, interval=60):
"""Retry instance launch until capacity is available.
Default: retry every 60s for up to 12 hours (720 attempts).
Always Free ARM shapes have intermittent capacity.
"""for attempt inrange(1, max_retries + 1):
try:
response = compute_client.launch_instance(launch_details)
print(f"Attempt {attempt}: SUCCESS — {response.data.id}")
return response
except oci.exceptions.ServiceError as e:
if e.status == 500and"Out of host capacity"instr(e.message):
jitter = random.uniform(0, 15)
print(f"Attempt {attempt}: Out of capacity. Retrying in {interval + jitter:.0f}s...")
time.sleep(interval + jitter)
else:
raise# Re-raise non-capacity errorsraise RuntimeError(f"Failed after {max_retries} attempts — no capacity available")
# ARM Always Free shape
arm_details = oci.core.models.LaunchInstanceDetails(
compartment_id=config["tenancy"],
availability_domain="Uocm:US-ASHBURN-AD-1",
display_name="hello-arm",
shape="VM.Standard.A1.Flex",
shape_config=oci.core.models.LaunchInstanceShapeConfigDetails(
ocpus=4, memory_in_gbs=24# Always Free max: 4 OCPUs, 24 GB
),
source_details=oci.core.models.InstanceSourceViaImageDetails(
image_id="ocid1.image.oc1.iad.aaaa...",
boot_volume_size_in_gbs=100# Always Free: up to 200 GB total
),
create_vnic_details=oci.core.models.CreateVnicDetails(
subnet_id="ocid1.subnet.oc1.iad.aaaa..."
),
metadata={
"ssh_authorized_keys": open(
os.path.expanduser("~/.ssh/id_rsa.pub")
).read()
}
)
launch_with_retry(compute, arm_details)
Step 6: Instance Lifecycle Operations
# Stop an instance (preserves boot volume)
compute.instance_action(instance_id=instance_id, action="STOP")
# Start a stopped instance
compute.instance_action(instance_id=instance_id, action="START")
# Reboot
compute.instance_action(instance_id=instance_id, action="RESET")
# Terminate (deletes instance, boot volume preserved by default)
compute.terminate_instance(instance_id=instance_id)
After launching your first instance, proceed to oraclecloud-local-dev-loop to set up a productive CLI-based workflow, or see oraclecloud-sdk-patterns for production-grade client lifecycle patterns.
