Configure Classic Multi-Attach Block Storage with OCFS2

Introduction

When using Classic Multi-Attach Block Storage, multiple instances can read and write to the same volume simultaneously. Standard filesystems such as EXT4 or XFS are not designed for concurrent multi-node access and can lead to data corruption in this scenario.

OCFS2 (Oracle Cluster File System 2) is a cluster-aware filesystem that coordinates concurrent writes across nodes, making it a suitable choice for Classic Multi-Attach volumes.

This guide walks through creating a 3-node OCFS2 cluster backed by a single classic-multiattach block volume in a 3AZ Public Cloud region.

Requirements

• An OpenStack client installed and configured
• An OVHcloud Public Cloud project in a 3AZ region

Architecture overview

The setup consists of 3 instances — one per availability zone — all attached to the same multi-attach volume, connected via a shared private network.

3AZ Public Cloud Region
┌──────────────────────────────────────────────────────────────┐
│  Zone A           │  Zone B           │  Zone C              │
│  ┌─────────────┐  │  ┌─────────────┐  │  ┌─────────────┐     │
│  │ Instance A  │  │  │ Instance B  │  │  │ Instance C  │     │
│  └──────┬──────┘  │  └──────┬──────┘  │  └──────┬──────┘     │
│         │         │         │         │         │            │
│  ┌──────┴─────────┴─────────┴─────────┴─────────┴─────────┐  │
│  │              Private Network (OCFS2 cluster)           │  │
│  └─────────────────────────┬──────────────────────────────┘  │
│                            │                                 │
│                ┌───────────┴────────────┐                    │
│                │  Block Volume          │                    │
│                │  (classic-multiattach) │                    │
│                └────────────────────────┘                    │
│  ┌──────────┐                                                │
│  │ Gateway  │  (internet access)                             │
│  └──────────┘                                                │
└──────────────────────────────────────────────────────────────┘

Create the infrastructure

Create a private network and subnet

openstack network create \
  --provider-network-type vrack \
  --provider-segment 10 \
  classic-multiattach-private-network

openstack subnet create \
  --dhcp \
  --network classic-multiattach-private-network \
  --subnet-range 10.1.0.0/16 \
  --dns-nameserver 213.186.33.99 \
  classic-multiattach-private-subnet

Create a gateway

openstack router create --external-gateway Ext-Net classic-multiattach-gateway
openstack router add subnet classic-multiattach-gateway classic-multiattach-private-subnet

Create the multi-attach block volume

openstack volume create \
  --size 10 \
  --type classic-multiattach \
  classic-multiattach-block-volume

Create an SSH keypair (skip if you already have one)

openstack keypair create --public-key ~/.ssh/id_rsa.pub classic-multiattach-keypair

Create three instances, one per availability zone

for i in {a..c}; do
  openstack server create \
    --flavor b3-8 \
    --image 'Ubuntu 24.04' \
    --network classic-multiattach-private-network \
    --key-name classic-multiattach-keypair \
    --availability-zone $(echo $OS_REGION_NAME | tr '[:upper:]' '[:lower:]')-$i \
    classic-multiattach-instance-$i
done

Attach the volume to all three instances

for i in {a..c}; do
  openstack server add volume classic-multiattach-instance-$i classic-multiattach-block-volume
done

Attach a floating IP to instance-a (for SSH access)

openstack floating ip create Ext-Net
openstack server add floating ip classic-multiattach-instance-a <YOUR_FLOATING_IP>

Retrieve the private IP of each instance

for i in {a..c}; do
  echo -n "classic-multiattach-instance-$i: " && \
  openstack server show classic-multiattach-instance-$i -c addresses -f json \
    | jq -r '.addresses."classic-multiattach-private-network"[0]'
done

Note down the three private IPs — you will need them in the OCFS2 configuration.

Configure OCFS2

The following steps must be performed on all three instances. Start with instance-a (via floating IP), then SSH from instance-a to instance-b and instance-c using agent forwarding.

SSH to instance-a

ssh -A ubuntu@<YOUR_FLOATING_IP>

On each instance: install OCFS2

sudo apt-get update
sudo apt-get install ocfs2-tools linux-modules-extra-$(uname -r)

On each instance: create the cluster configuration

sudo vim /etc/ocfs2/cluster.conf

cluster:
  name = ocfs2
  heartbeat_mode = local
  node_count = 3

node:
  number = 1
  name = classic-multiattach-instance-a
  ip_address = <YOUR_INSTANCE_A_PRIVATE_IP>
  ip_port = 7777
  cluster = ocfs2

node:
  number = 2
  name = classic-multiattach-instance-b
  ip_address = <YOUR_INSTANCE_B_PRIVATE_IP>
  ip_port = 7777
  cluster = ocfs2

node:
  number = 3
  name = classic-multiattach-instance-c
  ip_address = <YOUR_INSTANCE_C_PRIVATE_IP>
  ip_port = 7777
  cluster = ocfs2

On each instance: enable the O2CB cluster service

sudo sed -i 's/O2CB_ENABLED=false/O2CB_ENABLED=true/' /etc/default/o2cb
sudo systemctl restart o2cb

On instance-a only: format the volume with OCFS2

This step must be performed once only on one node.

The -N option sets the maximum number of node slots for the filesystem. This value can be increased later with tunefs.ocfs2, but it can never be reduced, so it is set here to match the 3 nodes used in this cluster.

sudo mkfs.ocfs2 -N 3 /dev/sdb

On each instance: mount the filesystem

sudo mount.ocfs2 /dev/sdb /mnt
sudo -s
echo "/dev/sdb /mnt ocfs2 _netdev,defaults 0 0" >> /etc/fstab
exit

Verify the cluster

On instance-a, write a test file:

cd /mnt
sudo touch test

SSH to instance-b from instance-a:

ssh ubuntu@<YOUR_INSTANCE_B_PRIVATE_IP>

Verify the file is visible:

ls -la /mnt/test

If the file appears, the OCFS2 cluster is functioning correctly and the shared volume is accessible from all nodes.

Go further

• For automated deployment using Terraform, see the public-cloud-examples repository
• Proper usage and limitations of Classic Multi-Attach Block Storage

Join our community of users.