I think I proposed the shared network namespace and ip per pod along with the per container filesystem/chroot. It might have been Tim though. Lots of bikeshedding to come up with the name Pod.
-
Another thing -- at the time Borg generally didn't use linux namespaces. As such the network was shared across the node, not across the alloc/container. Ports were assigned to containers and they were kept honest about which ports they were allowed to use.
Yes, dynamic port allocation had pervasive impact, on scheduling, configuration, discovery, monitoring, health checking, load balancing, various proxies, authentication, network isolation, and probably other things.
-
-
This is the general pattern though, and I'm waiting for it to hit k8s. You start out by keeping a port map (google did the same), and at some point hit a wall that you can't maintain that list anymore and need something more dynamic.
-
Dynamic ports changed everything because they were a massive shift from the way systems were designed up until that point. To be fair, most systems don't require that sort of scale still. However maintaining that list is still a PITA and should be offloaded to the scheduler.
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.