hmm. “stock” k8s still has limitations that make it look like a toy (how many nodes can you have in a cluster again? is that hyperscale?). the pattern I’ve see

hmm. “stock” k8s still has limitations that make it look like a toy (how many nodes can you have in a cluster again? is that hyperscale?). the pattern I’ve seen is that teams/companies that go into k8s hoping it’s going to solve the problems they have replace some of their problems (which k8s does address) with the operational burden of keeping k8s up-to-date and stable. the struggle to understand new things being rolled out and keeping up with the tech is real (another way of saying this is that k8s is nowhere close to having a story around deployment and maintenance that would make it easy to operate - big players can figure it out, but chances are you are not a big player and the big players want you to pay for their shit - pretty much like Rackspace wanted you to pay for openstack) another aspect that people don’t seem to get is that k8s is what google is doing (or used to do 5 years ago) internally but without the supporting infrastructure/experience that google has internally.