Just recently found out about auto_stop_machines = “suspend” (been using “stop” all this time)
Hoping to get some clarity on this. Is there any reason to use stop over suspend? Is there a cost benefit to truly scaling to zero? Or is the differential so insignificant that it wouldn’t matter?
I guess can anyone give me example of when you might use “stop” versus “suspend”?
The clock skew is the one that bites people the hardest, but it looks like that’s being actively worked on now. (Thanks, guys!)
As of today, though, I’m still seeing lag in my own tests, even with freshly created Machines. Possibly the patch hasn’t propagated through the entire fleet yet, …