Matthew Martin

Hot take: No RDS instance should ever be part of a #cloudformation stack.

The technology is so immature that it is just a matter of time before the stack gets in an unrecoverable state.

Everyone knows this, this is why they suggest splitting stacks into persistent and ephemeral, but really listen to yourself. If it is too dangerous to put RDS into a stack that could need to be dropped, it shouldn't be in *any* stack.

Feb 11 at 2:26pmElk
Show threadMatthew MartinFeb 11
This is a low-availability pattern by using a technology that has configuration complexity raise to the point that failure is guaranteed if you just wait long enough.
Show threadMatthew MartinFeb 11
This is a replay of SQL Server 2000 High availability clusters where the config was so complex and error prone it was just asking for downtime, except the vendor could blame it on the customer.
Show threadItamar Turner-TrauringFeb 11
@mistersql A good place to recommend this whole research program: https://tianyin.github.io/pub/csur.pdf