AWS ParallelCluster

3.15.0 verified Mon Apr 27 auth: no python

AWS ParallelCluster is an AWS supported open-source cluster management tool that enables you to deploy and manage high-performance computing (HPC) clusters on AWS. It automates the creation of HPC clusters with Slurm, AWS Batch, or other schedulers. The current version is 3.15.0, with releases roughly every 1-2 months.

pip install aws-parallelcluster

Common errors

error ModuleNotFoundError: No module named 'aws_parallelcluster' ↓

cause The correct module name is 'pcluster', not 'aws_parallelcluster'.

fix

Install with pip install aws-parallelcluster and import as from pcluster import pcluster.

error pcluster: command not found ↓

cause The ParallelCluster CLI is not installed or not in PATH.

fix

Run pip install aws-parallelcluster and ensure the Python Scripts directory is in PATH. Alternatively, use python -m pcluster.

Warnings

breaking In version 3.x, the CLI and API structure changed significantly from 2.x. Old scripts using 'pcluster' commands may fail. ↓

fix Migrate to ParallelCluster 3.x by updating commands and configuration files. Refer to the official migration guide.

deprecated Ubuntu 20.04 support is deprecated and removed after 3.13.0. ↓

fix Use Ubuntu 22.04 or later for new clusters.

gotcha The Python SDK (pcluster) is separate from the CLI. Importing 'pcluster' directly may not provide all CLI functionality. ↓

fix Use the CLI commands (e.g., `pcluster create-cluster`) for cluster operations, or use the SDK for programmatic access.

Imports

pcluster
wrong
```
import aws_parallelcluster
```
correct
```
from pcluster import pcluster
```
The correct module name is 'pcluster', not 'aws_parallelcluster'.

Quickstart

Quickstart for using ParallelCluster Python SDK to list clusters. Ensure AWS credentials are configured (e.g., via AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY).

import os
from pcluster import pcluster
client = pcluster.Client()
# Note: Requires AWS credentials configured via environment variables or AWS CLI.
# Example: list clusters
clusters = client.list_clusters()
print(clusters)