On Fri, Sep 15, 2017 at 3:04 PM, Alan Griffiths <apgriffiths79@gmail.com> wrote:

Hi,

Currently testing my recovery plan for HE failure.

I run a hosted engine, but in the event of some failure of the engine VM i want to be able to quickly restore the engine to a standalone host running independent of Ovirt (in this case a VM running on ESXi).

On a dry run I restore the backup onto the ESXi VM and run engine-setup. At this point it fails with an error that I am not in Global Maintenance mode.

How do you backup, and how do you restore?

If I put the cluster into maintenance mode before performing the backup then it works correctly. So is it necessary to do this on every backup? I can find nothing in the documentation detailing this as a requirement?

You might want to look at this: https://bugzilla.redhat.com/show_bug.cgi?id=1403903 and the linked patches.

Alternatively it seems I can manually update the ha_global_maintenance column in the vds_statistics table before running engine-setup, but not sure of the implications of doing this.

The way this works with engine-backup is: We added a new option DbJustRestored, which defaults to False. engine-backup --mode=restore sets it to True. engine-setup checks if it's False, and if so, goes on to check ha_global_maintenance. So in principle, it should be enough to: UPDATE vdc_options SET option_value='1' WHERE option_name='DbJustRestored'; However, this isn't exposed to engine-config, and is not considered an "external/official API".

ovirt-4.0.6.3

Above is for 4.1. You might want to upgrade. Or just reset ha_global_maintenance, but I am not aware of anyone verifying this or analyzing the implications. Best, -- Didi