3 ms·

> "the only concession I would make is that they could all be in the same daemon (one executable), but that's actually worse at scale, and harder to debug -- al

> "the only concession I would make is that they could all be in the same daemon (one executable), but that's actually worse at scale, and harder to debug -- all of these services can produce a lot of logs." This is already a thing, sort of, called hyperkube: https://github.com/kubernetes/kubernetes/blob/master/cluster/images/hyperkube/README.md https://github.com/kubernetes/kubernetes/blob/master/cluster... The caveat is that each daemon has to be started separately still

hardwaresofton 8y ago
hyperkube was what i used when I got my first cluster up and running (this was when coreos still hosted kubernetes baremetal setup guides not just pointing everyone to tectonic), I remember it fondly :) I didn't include it due to that caveat, and the general feeling that the processes really are meant to be started separately. I haven't seen anyone try and run these processes with some sort of supervisor but it seems like that's not "the way" and wouldn't really even offer any benefits.