Prepare and validate the virtual machines

Prepare the nodes for Ceph

You’ll prepare all three virtual machines (VMs) with hostname resolution, synchronized clocks, and the tools that cephadm needs to run storage services. Connect from the AGI host, then run the preparation commands inside each VM. Finish by identifying the empty data disks that Ceph will use.

Check the virtual machines

You created three Ubuntu 26.04 VMs named ceph1, ceph2, and ceph3.

Ensure that each VM has the following resources:

• Four virtual CPUs
• 4 GB of memory
• A 20 GB operating-system disk
• A separate unused data disk of at least 10 GB
• An interface connected to agi-net

For this Learning Path, the example unused disk is /dev/vdb. If you want to use a different disk, substitute /dev/vdb in the subsequent commands with your preferred device name.

Record the addresses

Start from a terminal on the AGI host with the ~/.ssh/agi_ceph private key that you created earlier. Connect to ceph1 as the ubuntu user:

    

        
        
ssh -o ConnectTimeout=10 -i "${HOME}/.ssh/agi_ceph" ubuntu@192.168.100.11

    

After login, your shell is running inside ceph1. The prompt typically includes ubuntu@ceph1.

Run the following commands in that SSH session to display the hostname, architecture, and IPv4 addresses of the VM:

    

        
        
hostname
uname -m
uname -r
ip -4 -brief address show scope global

    

Confirm that the architecture is aarch64. The tested Ubuntu 26.04 guest kernel is 7.0.0-34-generic. RADOS Block Device (RBD) requires kernel support for aes256k credentials. Loading the RBD module alone doesn’t prove authentication compatibility. Record one stable IP address for each VM.

The examples use the following addresses:

Run exit to leave the VM SSH session and return to the AGI host.

Repeat the login and address checks for ceph2 and ceph3, using 192.168.100.12 and 192.168.100.13 respectively in the SSH command.

You can also open a separate AGI-host terminal for each SSH session.

Configure the hosts file

After recording all three addresses, connect to each VM again and add all three nodes to the /etc/hosts of that VM. Run the following block once inside each VM SSH session, replacing the example addresses if needed:

    

        
        
cat <<'EOF' | sudo tee -a /etc/hosts
192.168.100.11 ceph1
192.168.100.12 ceph2
192.168.100.13 ceph3
EOF

    

In the SSH session connected to ceph1, confirm that the other nodes resolve and respond:

    

        
        
getent hosts ceph2 ceph3
ping -c 2 ceph2
ping -c 2 ceph3

    

Install and confirm the Ceph prerequisites

You need the following prerequisites for cephadm:

• Python 3
• systemd
• A container runtime
• Time synchronization
• LVM2

Run the following commands on all three VMs:

    

        
        
sudo apt update
sudo apt install -y chrony lvm2 podman openssh-server curl
sudo systemctl enable --now chrony ssh

    

Confirm the remaining services and tools:

    

        
        
python3 --version
podman --version
systemctl is-active chrony
command -v lvm

    

The output is similar to:

    

        
        Python 3.14.4
podman version 5.7.0
active
/usr/sbin/lvm

        
    

Check time synchronization on each VM:

    

        
        
chronyc tracking

    

Wait for a valid reference source and Leap status: Normal.

Validate the data disks

On each node, inspect the candidate disk:

    

        
        
lsblk -o NAME,SIZE,TYPE,FSTYPE,MOUNTPOINTS /dev/vdb
sudo wipefs --no-act /dev/vdb

    

The expected output is:

    

        
        NAME SIZE TYPE FSTYPE MOUNTPOINTS
vdb   10G disk

        
    

The disk should have no partitions, file system, or mount point. wipefs --no-act must not report any signatures. Resolve any unexpected output before continuing.

What you’ve accomplished and what’s next

You’ve prepared three networked VMs and confirmed that each VM has an unused data disk.

Next, you’ll bootstrap Ceph on ceph1 and add the first OSD.