Azure Planned Maintenance Strategy
I have concerns in regards to the use-cases of Azure and how the current Planned Maintenance strategy impacts this.
There are many use-cases where Availability Sets are of no use and do not protect you from downtime. VDI Platforms for example require persistent sessions to Single VMs. No amount of redundant VMs in an Availability Set will alleviate the impact to End Users if their Desktop VM is rebooted during business hours.
The Citrix Netscaler has recently been released on the Azure Marketplace which in theory should open up further use-cases for Azure. But why would I pay thousands to utilize Netscaler VPX Licences on Azure if all my user sessions can simply be knocked off in the middle of the day? The nature of the Netscaler involves persistent user sessions. Even with a High Availability Load-Balanced Netscaler set, if 1 Netscaler VM is rebooted, all those active sessions will be impacted and End-Users will be forced to reconnect to the system.
AWS gives you the control for managing the restart of your Single VMs during planned maintenance events. And Single VM's on AWS have supported SLA's.
I believe the entire Azure Planned Maintenance strategy needs a good re-think. At the very least Single VM's should only be impacted outside of normal business hours for each region.
To enable customers to address the valid points raised above, Azure should:
1) Provide better API access to the planned maintenance process. Integrate with Service Bus. Give long notice periods so workloads can be moved between update domains at a time that suits us.
2) Allow a much longer duration between update domains. Hours not minutes.