Deployment Guide for Rocky Linux 9¶

Tip

This tutorial is primarily designed for Rocky Linux 9, but it should be compatible with any RHEL-based distribution (e.g., Rocky Linux 8, AlmaLinux).

The instructions are tailored for the x86-64 architecture. For other architectures, such as ARM64, ensure you modify the download links and commands as needed.

Please run all commands as the root user throughout this tutorial.

1. Environment Preparation¶

1.1 Add EPEL Repository¶

dnf install -y yum-utils epel-release
dnf config-manager --set-enabled crb

1.2 Enable Time Synchronization¶

dnf install -y chrony
systemctl restart systemd-timedated
timedatectl set-timezone Asia/Shanghai
timedatectl set-ntp true

1.3 Configure Firewall¶

Tip

If you have multiple nodes, perform this step on each node. Otherwise, inter-node communication will fail.

Please see the config file /etc/crane/config.yaml for port configuration details.

systemctl stop firewalld
systemctl disable firewalld

If your cluster requires the firewall to remain active, open the following ports:

firewall-cmd --add-port=10013/tcp --permanent --zone=public
firewall-cmd --add-port=10012/tcp --permanent --zone=public
firewall-cmd --add-port=10011/tcp --permanent --zone=public
firewall-cmd --add-port=10010/tcp --permanent --zone=public
firewall-cmd --add-port=873/tcp --permanent --zone=public
firewall-cmd --reload

1.4 Disable SELinux (Optional)¶

# Temporary (will be reset after reboot)
setenforce 0

# Permanent (survives reboot)
sed -i s#SELINUX=enforcing#SELINUX=disabled# /etc/selinux/config

1.5 Select CGroup Version (Optional)¶

Rocky 9 uses CGroup v2 by default.

CraneSched supports both CGroup v1 and CGroup v2. However, using GRES on a CGroup v2 system requires additional configuration; see the eBPF guide for details.

1.5.1 Configure CGroup v1¶

If you cannot build the eBPF components and still need GRES, you can switch back to CGroup v1:

# Set kernel boot parameters to switch to CGroup v1
grubby --update-kernel=/boot/vmlinuz-$(uname -r) \
  --args="systemd.unified_cgroup_hierarchy=0 systemd.legacy_systemd_cgroup_controller"

# Reboot to apply changes
reboot

# Verify version
mount | grep cgroup

1.5.2 Configure CGroup v2¶

# Check if sub-cgroups have access to resources (expect to see cpu, io, memory, etc.)
cat /sys/fs/cgroup/cgroup.subtree_control

# Grant resource permissions to subgroups
echo '+cpuset +cpu +io +memory +pids' > /sys/fs/cgroup/cgroup.subtree_control

As mentioned earlier, using GRES with CGroup v2 requires the additional steps described in the eBPF guide.

2. Install Toolchain¶

The toolchain must meet the following version requirements:

CMake ≥ 3.24
If using clang++, version ≥ 19
If using g++, version ≥ 14

Use the following commands to install and enable the required toolchain:

dnf install -y \
    gcc-toolset-14 \
    cmake \
    patch \
    flex \
    bison \
    automake \
    ninja-build

echo 'source /opt/rh/gcc-toolset-14/enable' >> /etc/profile.d/extra.sh
source /etc/profile.d/extra.sh

3. Install Project Dependencies¶

dnf install -y \
    libstdc++-devel \
    libstdc++-static \
    openssl-devel \
    curl-devel \
    pam-devel \
    zlib-devel \
    libaio-devel \
    systemd-devel \
    libcurl-devel \
    elfutils-libelf-devel \
    shadow-utils-subid-devel

4. Install and Configure MongoDB¶

MongoDB is required on the control node only.

Please follow the Database Configuration Guide for detailed instructions.

5. Install and Configure CraneSched¶

5.1 Build and Install¶

Configure and build CraneSched:

git clone https://github.com/PKUHPC/CraneSched.git
cd CraneSched
mkdir -p build && cd build

# For CGroup v1
cmake -G Ninja ..
cmake --build .

# For CGroup v2
cmake -G Ninja .. -DCRANE_ENABLE_CGROUP_V2=true
cmake --build .

Install the built binaries:

Tip

We recommend deploying CraneSched using RPM packages. See the Packaging Guide for installation instructions.

cmake --install .

For deploying CraneSched to multiple nodes, please follow the Multi-node Deployment Guide.

5.2 Configure PAM Module¶

PAM module configuration is optional but recommended for production clusters to control user access.

Please follow the PAM Module Configuration Guide for detailed instructions.

5.3 Configure the Cluster¶

For cluster configuration details, see the Cluster Configuration Guide.

6. Start CraneSched¶

Using systemd (Recommended)¶

Control node only: Create crane user (automatic with RPM packages):

sudo groupadd --system crane 2>/dev/null || true
sudo useradd --system --gid crane --shell /usr/sbin/nologin --create-home crane 2>/dev/null || true

Then start services:

systemctl daemon-reload
systemctl enable cranectld --now  # Control node
systemctl enable craned --now     # Compute node

Running manually (foreground)¶

cranectld  # Control node
craned     # Compute node