I've said it before and I'll say it again: never u...
# random
h
I've said it before and I'll say it again: never use t3 instances on AWS
😮 1
w
Do tell....
h
Once you go over the CPU overdraft limit, the instance turns into jelly and your whole production environment falls over
BUT THE INSTANCE STAYS UP
and reports itself as healthy
basically you inevitably end up bursting too often
Just go with an m5
w
Whoaaa
That's whack
h
They are cheaper, which is why people use them I guess
but you are on borrowed time
literally
c
That was the default a while back. These days the burstable instances default to unlimited CPU burst. https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/burstable-performance-instances-unlimited-mode.html
g
Every time I hear about AWS I'm happy to only work with GCP... 😄
c
Last I looked a few years back the performance of a burstable instance is equivalent to a non-burstable, and the burst charges make it cost essentially the same as the equivalent non-burstable. So for workloads that those instances are suitable, it's financially a no-brainer to use burstable instances with unlimited burst config
A lot of AWS problems are due to their philosophy of supporting all features since the initial release of a service (even APIs!). It's getting to the point of the Microsoft legacy support
Maybe by itself it's not a bad philosophy, but coupled with the drive to be first to release a new service results in some not particularly fully thought out features
GCP might have fewer services but they seem better planned
g
Heh. Compared to the deathmarch of Google products I guess neither is optimal.
c
Google free products definitely follow a throw-at-wall and see what can be monetized later, but GCP and Google Workplace have been stable. If anything, slow to innovate.