Trying to Understand What Triggers Autoscaling

[Anonymous]

We had an issue a few months back where we got the out of memory error and so we turned on autoscale to avoid that while we got a handle on our usage. So we have been using the Fabric Capacity Metrics App to monitor our usage but there does not seem to be a clear link from what we see in our utlilization and what is triggering the occasional autoscaling to happen. You can see we are over 100% of our CUs daily but Autoscalse has only kicked in three times in 3 months, this morning being one of them. And I do see our background utilization was up this morning, but that is still not giving me any detail on what is triggering us to autoscale and what reports to look at for optimization. Any advice would be appreciated.

[SaiTejaTalasila]

Hi @Anonymous ,

Go to your system events visual in fabric capacity metrics app it will show the history of interactive delay and how many minutes overload occurred(the difference b/w overloaded and active). If you don't have autoscale enabled and the interactive operations are not going down then your end-users will see loading issues or rendering issues while they interact with the reports.

Thanks,

Sai Teja 

[lbendlin · Accepted solution]

10 minute - 60 minute - auto scale

up to 10 minute of overconsumption: no penalty

up to 60 minutes of overconsumption: interactive delay

after 60 minutes - autoscale if allowed, or total capacity lockdown if not.

Autoscale will be active for 24 hrs, if you need it or not.

[Anonymous]

Thank you, that actually makes sense. So I am correctly identify the items that are using the most capacity and we are just not usually over capacity for long enough to trigger.