A.11. VDSM Hook Execution

before_vm_start scripts can edit the domain XML in order to change VDSM’s definition of a virtual machine before it reaches libvirt. Caution must be exercised in doing so. Hook scripts have the potential to disrupt the operation of VDSM, and buggy scripts can result in outages to the Red Hat Virtualization environment. In particular, ensure you never change the UUID of the domain, and do not attempt to remove a device from the domain without sufficient background knowledge.

Both before_vdsm_start and after_vdsm_stop hook scripts are run as the root user. Other hook scripts that require root access to the system must be written to use the sudo command for privilege escalation. To support this the /etc/sudoers must be updated to allow the vdsm user to use sudo without reentering a password. This is required as hook scripts are executed non-interactively.

Example A.4. Configuring sudo for VDSM Hooks

In this example the sudo command will be configured to allow the vdsm user to run the /bin/chown command as root.

  1. Log into the virtualization host as root.
  2. Open the /etc/sudoers file in a text editor.
  3. Add this line to the file:

    vdsm ALL=(ALL) NOPASSWD: /bin/chown

    This specifies that the vdsm user has the ability to run the /bin/chown command as the root user. The NOPASSWD parameter indicates that the user will not be prompted to enter their password when calling sudo.

Once this configuration change has been made VDSM hooks are able to use the sudo command to run /bin/chown as root. This Python code uses sudo to execute /bin/chown as root on the file /my_file.

retcode = subprocess.call( ["/usr/bin/sudo", "/bin/chown", "root", "/my_file"] )

The standard error stream of hook scripts is collected in VDSM’s log. This information is used to debug hook scripts.