STONITH in Microsoft Azure Environments

The Azure CLI is a command line interface supported by Microsoft to manage Azure resources such as virtual machines. Used with STONITH, it allows the cluster software to power off a cluster node that appears to have died thus ensuring that the unhealthy node cannot access or corrupt any shared data.

Requirements

Package

Azure CLI – install Linux Azure CLI on each server in the cluster.

Custom Role

The virtual machines on Microsoft Azure used for Azure Fencing and the custom roles assigned to users must have at least powerOff permissions on the virtual machines.

• Microsoft.Compute/*/read
• Microsoft.Compute/virtualMachines/powerOff/action
• Microsoft.Compute/virtualMachines/start/action

Pre-checking

On each server run the following command to verify that authentication on Microsoft Azure is working.

# az vm show --resource-group <group name> --name <vm name>

STONITH Installation

Install the LifeKeeper for Microsoft Azure STONITH script on each server where LifeKeeper is installed and communication paths are configured to all servers by running the following command:

# /opt/LifeKeeper/samples/STONITH/azure-stonith-install

Since the above command works interactively, enter the group name and check the virtual machine name on Microsoft Azure for the cluster node displayed.

Example output from the command

STONITH script install...
Please enter the Resource Group name in Azure: rg-Group1
　
Please enter the System name in Azure[vm-HostA]:
Enable Stonith on node vm-HostA [Yes]:
s
Please enter the System name in Azure[vm-HostB]:
Enable Stonith on node vm-HostB [Yes]:
Configuration file /opt/LifeKeeper/config/stonith.conf was saved.

Update /opt/LifeKeeper/config/stonith.conf

After the installation is completed, the settings to power off the virtual machine will be added to the following file.

/opt/LifeKeeper/config/stonith.conf

LifeKeeper STONITH configuration
　
Example: <host> az vm restart -g <resource group> -n <node name>
vm-HostA az vm stop -g rg-Group1 -n vm-HostA --skip-shutdown
vm-HostB az vm stop -g rg-Group1 -n vm-HostB --skip-shutdown

*Note: ‘stop’ is recommended over restart to avoid fence loops (i.e. two machines have lost communication but can still STONITH each other, taking turns powering each other off and rebooting). ‘skip-shutdown’ is recommended to quickly power off the node.