Configuration

This page is the exact lookup for where and how a Job or Server runs. For decision guidance, start with GPUs & VM Types, Cloud Credentials, and Buckets.

Configuration mappings

Purpose	CLI	Python SDK	Notes
Pinned compute credential	--credentials NAME	CloudConfig(credentials=...) or Client(credentials=...)	Omit for an eligible saved-credential pool
Region	--region REGION	CloudConfig(region=...)	Omit for normal multi-region selection
GPU model and count	--gpu-type TYPE	submit(gpu=...), decorator gpu=...	Submit-level value, not a CloudConfig field
Exact VM	--vm-type TYPE	CloudConfig(vm_type=...)	Mutually exclusive with GPU type
Spot	--spot	CloudConfig(spot=True)	Jobs only
Local compute	--local	CloudConfig(cloud_provider=CloudType.Local)	Requires local Docker capability on the API
Root disk capacity	--disk-size GB	CloudConfig(disk_size_gb=...)	Independent of performance tier
Root disk tier	--disk-tier TIER	CloudConfig(disk_tier=...)	medium, high, or ultra
Availability zone	--zone ZONE	CloudConfig(availability_zone=...)	Provider-specific value
Input Bucket	--input-bucket NAME	input=Bucket(...) or input_bucket	Jobs only; must exist
Output Bucket	--output-bucket NAME	output=Bucket(...) or output_bucket	Jobs only; created if needed
Environment	-e, --env-file	env={...}	Explicit values override Secret keys
Named Secret	--secret NAME	secrets=[...]	Repeatable
Container runtime	Docker flags below	docker_options={...}	Applies after hardware resolution

Credential precedence

Python resolves compute credentials in this order:

1. An explicit credential object in CloudConfig
2. A saved credential name passed to CloudConfig or Client
3. The sole saved credential, when exactly one exists

from anycloud.types import AWSCredentials, CloudConfig

explicit = CloudConfig(
    credentials=AWSCredentials(
        accessKeyId="AKIA...",
        secretAccessKey="...",
    ),
)

saved = CloudConfig(credentials="production-aws")

A CLI --credentials value wins over the ANYCLOUD_CREDENTIALS_NAME environment fallback. If the CLI omits both, the server can select among eligible saved named credentials; an unpinned compute configuration must also leave region unpinned.

The Python SDK does not guess when multiple saved credentials exist. Pass Client(credentials="name") or an explicit CloudConfig.

If a pinned credential fails authentication, the deployment becomes invalid. With unpinned compute, Anycloud can temporarily block the failed credential and select another compatible saved credential.

Hardware constraints

gpu / --gpu-type and vm_type / --vm-type are mutually exclusive.

• GPU selection accepts a model with optional count, such as h100 or a100:8.
• VM selection accepts a provider-specific instance type, such as g6e.xlarge, Standard_NC40ads_H100_v5, or n1-standard-8.
• Repeating a CLI GPU or VM flag creates an ordered fallback pool. Python SDK submit parameters currently accept one GPU or VM target per Submission.
• --gpus is a Docker exposure setting; it does not select the cloud VM.

# Ordered GPU preference
anycloud submit IMAGE \
  --gpu-type h100:8 \
  --gpu-type a100:8

# Exact VM
anycloud submit IMAGE \
  --credentials production-aws \
  --vm-type g6e.xlarge

Region pinning disables regional failover. Use it only for residency, debugging, quota work, or another explicit constraint.

CloudConfig fields

Python field	Type	Description
credentials	str | CloudCredentials	Saved name or explicit credential object
cloud_provider	CloudType	Inferred from credentials when omitted
vm_type	str	Exact VM type
spot	bool	Spot/preemptible capacity; Jobs only
region	str	Pinned compute region
availability_zone	str	Provider availability zone
disk_size_gb	int	Root OS disk capacity
disk_tier	DiskTier	medium, high, or ultra
input_bucket	str	Read-only Job input at /mnt/input
output_bucket	str	Job output at /mnt/output
input_storage_credentials	CloudCredentials	Cross-account input identity
input_storage_region	str	Cross-account input region
output_storage_credentials	CloudCredentials	Cross-account output identity
output_storage_region	str	Cross-account output region
checkpoint_storage_credentials	CloudCredentials	Cross-account checkpoint identity
checkpoint_storage_region	str	Cross-account checkpoint region

GPU selection is deliberately absent: pass gpu= to Client.submit(), @anycloud.function(), or @anycloud.serve().

Configuration inheritance

A client-level CloudConfig supplies defaults for its deployments. A deployment-level configuration takes precedence.

CloudConfig.replace() returns a new configuration with selected changes:

from anycloud.types import CloudConfig

base = CloudConfig(
    credentials="production-aws",
    spot=True,
    disk_size_gb=400,
    disk_tier="high",
)

on_demand = base.replace(spot=False)
west = base.replace(region="us-west-2")

Submission.replace() provides the same immutable-variant pattern for submit_many().

Bucket credential rules

Input, output, and checkpoint storage credentials are independent. When a storage credential is omitted, that Bucket uses compute credentials; setting one Bucket's storage credential does not affect the others.

On the CLI, cross-account storage requires a credential and region:

anycloud submit IMAGE \
  --credentials compute-account \
  --input-bucket shared-data \
  --input-storage-credentials storage-account \
  --input-storage-region us-east-1

In Python, a Bucket sourced from another Client carries the storage identity:

compute = anycloud.Client(credentials="gcp-compute")
storage = anycloud.Client(credentials="aws-storage")

job = compute.submit("IMAGE", input=storage.bucket("shared-data"))

Cross-cloud storage currently supports S3. It uses scoped STS credentials that last up to 36 hours and are not refreshed mid-run.

Root disk tiers

Disk capacity and performance are independent. --disk-size 100 --disk-tier high means a 100 GB root disk at the high tier. medium is the default.

On AWS, tiers map to gp3 root-volume settings:

Tier	IOPS	Throughput
medium	3,000	125 MiB/s
high	8,000	1,000 MiB/s
ultra	16,000	1,000 MiB/s

On Azure, medium uses Standard storage; high and ultra use Premium SSD. Azure currently maps ultra to the same Premium tier as high. The VM size must support Premium storage.

Disk tiers are supported on AWS and Azure. Other providers reject a non-default tier.

Docker runtime options

CLI	docker_options key	Example
--gpus	gpus	all, 2
--shm-size	shmSize	8g
--memory	memory	32g
--cpus	cpus	4
--ipc	ipc	host
--ulimit	ulimits	repeatable
--runtime	runtime	nvidia
--bind	binds	repeatable host bind

Anycloud configures GPU access automatically for GPU-backed VMs. --gpus controls what Docker exposes; it is not normally necessary to specify an NVIDIA runtime.

Runtime environment

Every workload receives:

Variable	Meaning
DEPLOYMENT_ID	The deployment's unique ID
PORT	Server listen port; defaults to 8088

Jobs with Buckets also receive the fixed mount paths /mnt/input, /mnt/output, and /mnt/checkpoint. See Buckets.