With the maintenance operation, you can schedule recurring or immediate maintenance of all the RDS hosts in an automated instant-clone farm. During each maintenance cycle, all the RDS hosts are refreshed from the golden image virtual machine.
You can make changes to the golden image virtual machine without affecting the RDS host instant clones because the snapshot of the current golden image VM is used for maintenance. The instant clones created in the automated farm use the information in the golden image VM for their system configuration.
You can schedule maintenance on an automated farm but not on individual RDS hosts in the farm.
If possible, schedule maintenance operations during off-peak hours to ensure all that RDS hosts have finished maintenance and are available during peak hours.
Prerequisites
- Decide when to schedule the maintenance operation. By default, Connection Server starts the operation immediately.
You can schedule an immediate maintenance or recurring maintenance or both for a farm. You can schedule maintenance operations on multiple farms concurrently.
- Decide whether to force all users to log off when the maintenance operation begins or wait for each user to log off before refreshing that user's machine.
If you force users to log off, Horizon notifies users before they are disconnected and allows them to close their applications and log off.
- Decide the minimum farm size. The minimum farm size is the number of RDS hosts that are kept available at all times to allow users to continue to use the farm. For example, if the farm size is ten and the minimum farm size is two, then maintenance will be performed on eight RDS hosts. As each RDS host becomes available again then the remaining hosts will go through maintenance. All RDS hosts are managed individually, so as one host becomes available then one of the remaining hosts will be put into maintenance.
However, if you schedule immediate maintenance, then all the RDS hosts in the farm will be put into maintenance.
All RDS hosts will also be subject to policy and will wait for logoff or force users to logoff depending upon what policy is configured.
- Decide whether to stop provisioning at first error. If you select this option and an error occurs when Connection Server provisions an instant-clone, provisioning stops. You can select this option to ensure that resources such as storage are not consumed unnecessarily.
Selecting the Stop at first error option does not affect customization. If a customization error occurs on an instant-clone, other clones continue to be provisioned and customized.
- Verify that provisioning is enabled. When provisioning is disabled, Horizon stops the machines from being customized after they are refreshed.
- If your deployment includes replicated Connection Server instances, verify that all instances are the same version.
