51–58 of 58 posts
Earlier quoted context omitted.
> I think all big cloud providers replaced etcd with a custom in house implementation in their K8s offerings for this reason This is incorrect. AWS, Azure, GCP still use etcd in their customer kubernetes services. "EKS Ultra Scale" uses an in-house Journal
GKE uses spanner (with an etcd API layer) per the docs. https://docs.cloud.google.com/kubernetes-engine/docs/concept... https://cloud.google.com/blog/products/containers-kubernetes...
This sounds like a very direct approach to linearizability - just put everything in a linear order. But this includes READ operations too! You have to get global consensus for every read! Most distributed systems have additional complexity and latency on write paths so that reads can be completely local. If you can accept slow read operations this seems like a great trade off, but I think that is going to relegate th…
The article says that if your read hits a master there’s no consensus needed. Perhaps I read wrong
Of course, you could build database that uses meerkat to do leader election, and then neither reads nor writes need to go through the consensus algo. But then you've basically just reproduced conventional distributed db, and you're back to the timeout problems this was supposed to solve.
Earlier quoted context omitted.
The article says that if your read hits a master there’s no consensus needed. Perhaps I read wrong
The system seems to have no master or leader at all. Where in the article did you get that reads from a master have no consensus? Of course, you could build database that uses meerkat to do leader election, and then neither reads nor writes need to go through the consensus algo. But then you've basically just reproduced conventional distributed db, and you're back to the timeout problems this was supposed to solve.
Earlier quoted context omitted.
The article says that if your read hits a master there’s no consensus needed. Perhaps I read wrong
The system seems to have no master or leader at all. Where in the article did you get that reads from a master have no consensus? Of course, you could build database that uses meerkat to do leader election, and then neither reads nor writes need to go through the consensus algo. But then you've basically just reproduced conventional distributed db, and you're back to the timeout problems this was supposed to solve.