AVD Scaling plans are too simplistic

AVD Scaling plans are too simplistic
2

Upvotes

Upvote

 Jan 27 2022
5 Comments (5 New)
Tell us more

AVD scaling plans are missing some basic feature and are restricted to a single day schedules.

I'm missing:

 - option to do a dry run to see what would happen if I enable it

 - option for several messages in different intervals to the users about the shutdown

 - option to do only some of the tasks (for example: only notify and shutdown for personal desktops and only scale up for hostpools running over the weekend for after hour employees)
 - exclusion tag is applied to VM while drain mode is applied to session host object in host pool - this is not a good design and will lead to people forgetting about one or the other when doing maintenance and having to set the flags in two different places

 - plan is restricted to a single day, I can't start hostpool of Friday, scale over weekend and shutdown on Monday
 - option to ramp down at the end of the day and scale up to a smaller threshold over night

Comments
Microsoft

Hi @enecoTD , thank you for your feedback! I have a few follow up questions:

- For the option to do a dry run to see what would happen if I enable it, how would this work? Could you just enable Autoscale in a validation host pool to see that it works as expected before enabling it in a production environment?

 - For the option to send several messages in different intervals to the users about the shutdown, what is the use case for this?

 - For the option to ramp down at the end of the day and scale up to a smaller threshold over night, how is this different from the ramp down and off peak hour configuration flexibility that scaling plans afford today?

Microsoft
Status changed to: Tell us more
 
Copper Contributor

I'd like to upvote the exclusion tag part brought up by the OP.  Why can't drain mode be used as the parameter used to determine if a session host should be included in the scaling logic or not?  Having to put a tag in manually and the drain mode option often leads to issues.

 

We are trying to deal with this using pipelines to set tagging and drain mode for VMs for patching Cycles / when a session host needs maintenance but I can't help feel it would be a lot cleaner if existing options were used for the scaling logic.

 

Thanks,

Microsoft

HI everyone, The AVD scalling plans do have some issues, they don't work for many customers, I handled cases for. There is an issue with the avd schedule as it does not stop the VM at ramp down time and have to be manually stopped.

Moreover, the message popup before the Force stop does not appear.

We see that the customer open cases for the same and its a bit tricky for us engineers also to find the reason behind it.

They are confused about the threshold percentage in Ramp up and down as they find it difficult to understand the Calculation, leading to errors.

Forgive if I'm being too naive.

Microsoft

Hey @tejas_memane Thank you for the feedback! I would love to learn more about the issues that you described and we may need to troubleshoot the issue if you have customer cases to share.

Can you also help me to understand what is causing the confusion in the threshold settings so that I can look into whether we can improve the docs for better understanding? Autoscale service checks used host pool capacity (# of users sessions / # of host pool capacity) against the capacity threshold defined by the customer, 

  • If the used host pool capacity is below the capacity threshold and autoscale can turn off virtual machines (VMs) without going over the capacity threshold, then the feature will turn off the VMs.
  • If the used host pool capacity goes over the capacity threshold, then autoscale will turn on more VMs until the used host pool capacity goes below the capacity threshold.

We also have some example scenarios that can help understand how autoscale feature works with the defined capacity threshold.

Autoscale scaling plans and example scenarios in Azure Virtual Desktop | Microsoft Learn