I don’t think Google can divest GCP without losing a fatal amount of face. It would cause some very awkward questions to be asked such as: is Google’s valuation justified if they fail so abjectly at what is supposed to be their core competence?

On the other hand, from the beginning it has never felt like they wanted to be in the cloud services business and GCP has only ever existed for the aforementioned face saving reasons. It has just never fit with their company culture of: NIH all the tech, have users not customers, and place what’s best for managers’ careers above what’s best for the company.

They’re also disabling io_uring in some of their public facing services, chromebooks, etc. The issue is they run a bunch of kernels that are years old, which were missing security backports.

There has been one io_uring CVE for the linux 6.x series. That’s worse than zero, but local linux kernel exploits are pretty common.

Anyway, this sort of rug pulling for stuff like GKE makes it clear they’re not willing to stand by already-released products.

Edit: Also, it’s worth pointing out that the kernels for chromebooks and gke are single-tenant, so, if disabling io_uring really is necessary, it should be a customer controlled knob.

If people are running kernel exploits on your laptop / server containers, you’re probably already screwed.

Sorry, could you expand a bit more on what happened with GKE? I’m still trying to process all of this instead of enjoying my Saturday.

I'm a huge proponent of GCP - done both green field and AWS -> GCP migrations.

However, losing google cloud domains is hard to defend, and makes promoting GCP hard.

anecdotally ive seen schools buy into gcp but they are heavily invested in chromebooks and google apps for students