Skip to main content

Charmed HPC Localhost Deployment

Deploy production-like HPC clusters on your localhost using Charmed HPC, LXD containers, and Juju orchestration.

Overview​

Charmed HPC provides a complete High Performance Computing stack that can be deployed locally using:

  • LXD: Container hypervisor for lightweight virtualization
  • Juju: Application orchestration and lifecycle management
  • SLURM: Workload manager and job scheduler
  • Vantage CLI: Streamlined deployment and management

Prerequisites​

Before proceeding, ensure you have completed the Charmed HPC prerequisites from the localhost deployment applications overview.

Required Components​

  • LXD (latest/stable)
  • Juju (3.6/stable)
  • Vantage CLI
  • Bootstrap Juju controller on localhost

Available Applications​

SLURM Juju Localhost​

Deploy a complete SLURM cluster using Juju charms on LXD containers.

vantage app deploy slurm-juju-localhost

Features:

  • Multi-node SLURM cluster in containers
  • Head node with scheduler and controller
  • Compute nodes for job execution
  • Shared filesystem
  • User management integration

Architecture:

  • Controller: SLURM controller daemon (slurmctld)
  • Database: SLURM accounting database (slurmdbd)
  • Compute Nodes: SLURM compute daemon (slurmd)
  • Login Node: User access point

Quick Start​

1. Deploy SLURM Cluster​

# Deploy the SLURM cluster
vantage app deploy slurm-juju-localhost

# Monitor deployment progress
vantage app status slurm-juju-localhost

2. Access Your Cluster​

# Get cluster connection details
vantage app info slurm-juju-localhost

# Connect to login node
juju ssh slurm-login/0

3. Submit Test Jobs​

# Simple test job
srun --nodes=1 --ntasks=1 hostname

# Batch job example
sbatch <<EOF
#!/bin/bash
#SBATCH --job-name=test-job
#SBATCH --output=test-output.txt
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --time=00:05:00

echo "Hello from SLURM on localhost!"
hostname
date
EOF

Configuration​

Cluster Sizing​

Customize your cluster configuration:

# Deploy with specific node count
vantage app deploy slurm-juju-localhost --config nodes=4

# Scale compute nodes
vantage app scale slurm-juju-localhost --compute-nodes=6

Resource Allocation​

# Configure memory and CPU limits
vantage app config slurm-juju-localhost \
  --memory-per-node=4GB \
  --cpus-per-node=2

Management Commands​

Cluster Operations​

# Start cluster
vantage app start slurm-juju-localhost

# Stop cluster
vantage app stop slurm-juju-localhost

# Restart cluster
vantage app restart slurm-juju-localhost

# Delete cluster
vantage app delete slurm-juju-localhost

Monitoring​

# Check cluster status
vantage app status slurm-juju-localhost

# View logs
vantage app logs slurm-juju-localhost

# Get detailed information
vantage app describe slurm-juju-localhost

Troubleshooting​

Common Issues​

Deployment Failures:

# Check juju status
juju status

# View application logs
juju debug-log --include slurm-controller

# Retry failed deployments
vantage app deploy slurm-juju-localhost --force

Container Issues:

# List LXD containers
lxc list

# Check container logs
lxc logs <container-name>

# Restart containers
lxc restart <container-name>

Network Connectivity:

# Check LXD network
lxc network show lxdbr0

# Test container connectivity
lxc exec <container-name> -- ping google.com

Performance Tuning​

Container Resources:

# Increase container memory
lxc config set <container-name> limits.memory 4GB

# Set CPU limits
lxc config set <container-name> limits.cpu 2

SLURM Configuration:

# Access SLURM configuration
juju ssh slurm-controller/0
sudo nano /etc/slurm/slurm.conf

# Restart SLURM services
sudo systemctl restart slurmctld

Advanced Topics​

Custom Charms​

Deploy custom or development charms:

# Deploy from local charm
vantage app deploy slurm-juju-localhost --charm-path=./my-slurm-charm

# Use specific charm revision
vantage app deploy slurm-juju-localhost --charm-revision=25

Integration​

Storage Integration:

# Add shared storage
juju deploy ceph-osd --storage osd-devices=/dev/sdb
juju relate slurm-controller ceph-osd

Monitoring Integration:

# Add Prometheus monitoring
vantage app deploy prometheus-localhost
vantage app relate slurm-juju-localhost prometheus-localhost

Next Steps​

Support​

For issues specific to Charmed HPC localhost deployments: