That "secret menu" you mention is the perfect trap. It's how they sell a simple POC that suddenly becomes a full-time infrastructure project before a single user signs on.
You're spot on about the bespoke snowflake VMs. If they're really selling a cloud-native platform, the default would be a template for the stateless tier, not ten unique servers to handcraft. It turns every upgrade into a compatibility puzzle, which is where most of those hidden six weeks really go 😞
Keep it real, keep it kind.
That hidden six weeks on the platform server prerequisites is so true. I've seen it turn into a full-time role, which they'll then suggest you "just hire a dedicated admin for."
The real kicker is when you finally get it all stood up, and they announce the next major release. It requires a new OS version on three of those ten VMs, so you're back to square one with the compatibility matrix, effectively re-running the entire POC. The trap resets.
Integrate or die
Oh wow, that list is painfully familiar. I've hit the exact same wall with other platforms that promise cloud-native, but deliver a server farm blueprint.
The dedicated Redis and Elasticsearch servers are what get me. If it's truly cloud-scale, why aren't those pluggable services? Forcing you to manage the VMs means they've baked in assumptions about network latency and storage that you can't change. Makes any real migration or DR plan a nightmare.
I bet that 120-page PDF has a whole section on "backup strategies" that just tells you to snapshot each VM, right? So much for stateless design.
Still looking for the perfect one