Resource Provider

Compute Resource Provider (CRP) - > Virtual Machine Scale Sets (VMSS)

Description of Feature or Work Requested

VMSS Rolling Upgrade provides an upgrade facility that automatically applies VMSS model changes to the individual instances through a health-based orchestration.
There are some additional opt-in flags we are introducing for Upgrade policy for VMSS. At the moment we expose these four properties that our customers configure -

1. maxBatchInstancePercent - The maximum percent of total virtual machine instances that will be upgraded simultaneously by the rolling upgrade in one batch. Default: 20%
2. maxUnhealthyInstancePercent - The maximum percentage of the total virtual machine instances in the scale set that can be simultaneously unhealthy. Default: 20%
3. maxUnhealthyUpgradedInstancePercent - The maximum percentage of upgraded virtual machine instances that can be found to be in an unhealthy state. Default: 20%
4. pauseTimeBetweenBatches - The wait time between completing the update for all virtual machines in one batch and starting the next batch. Default: 0 seconds

In addition to these four, we are introducing two new properties. Note that these are opt-in properties.

1. enableCrossZoneUpgrade - Setting this Boolean property will allow VMSS to ignore AZ boundaries when constructing upgrade batches, and only consider Update Domain and maxBatchInstancePercent to determine the batch size.
2. prioritizeUnhealthyInstances - Setting this Boolean property will lead to all unhealthy instances in a scale set getting upgraded before any healthy instances.

Minimum API Version Required

2020-12-01

Swagger Link

Azure/azure-rest-api-specs#11184
Rolling upgrade - Prioritize Unhealthy VMs.docx

Target Date

2020-12-20