Unable to load model due to reaching capacity limits (For the past 5 hours??)

[GTS_ONE]

Hi All,

As of today, all users in our PBI Premium Capacity are receiving this message on the Service.  While the error itself seems clear, what's confusing me is that someone in London first reported this, so by the time we came online in the US, this message had been present for at least 5 hours.

Even if we hit the limit of the Capacity, it shouldn't lock up for this long, should it?

I'm also wondering if maybe there's something happening at the MS level that's affecting us and causing this.

I do have a critical support ticket open (still waiting for the callback).  

But just wondering if anyone has ever seen this message for an extended length of time (and any thoughts on how to clear the memory).  I believe you can pause/resume the Capacity but I'm not the admin so I've contacted them about that, too.

Thanks

[gp10]

Yes the new policy is pretty vague.
We had to wait several hours for our capacity to start working again after we enabled Autoscale.
We have no idea why a spike like that in CU usage happened from a dataflow's refresh.
This wasn't possible before.
I think its the new normal, you cant just have a frozen capacity like that. You need answers and options to prevent this. Why not drop the operation that causes this and instead kill the entire capacity?
Doesn't make sense.
Also, the documentation and support in such an important issue is amateurish from Microsoft.

[GTS_ONE]

Updating my own post.  So we opened a MS Support ticket and they are telling us that this is all due to the new throttling policy rolled out on Oct 1st where you "borrow" overloaded time.  And the overloaded time you spend today is borrowed from tomorrow's total.  And if you accumulate more than 24 hours of borrowed time, your Capacity just stops cold and nothing works.

But so much of this didn't add up when it comes to why we lost service today.

For one thing, the server only stops responding if you accumulate more than 24 hours of borrowed time.  But between 2 and 24 hours, filters and things stop working on your reports.  We saw absolutely nothing like that at all.  It's as if we somehow accumulated 24 hours of overage time within 8 hours of physical time.

Also, no one could explain the math behind how borrowed time gets paid back.  One Support engineer told us that now that we were locked out from spending more than 24 hours, we wouldn't get service back until tomorrow!  And yet service came back up about 8 hours after it shut down.

They said we could use the new Fabric Metrics app to check for this usage, but while the dashboards do show overload (just like the older versions), they simply don't give any clear metrics of where you stand on that 0 to 24 hour timeline.  And certainly there's no warning as you start to approach that 24th hour!

I'm genuinely hoping that something actually failed on our capacity server and that that's what caused all this.  Because if not, and this is just the new normal, then it is going to be a nightmare trying to administer everything.