Just wondering the reason for huge default-address-pool per container 172.x.0.0/16? This is going to chew through address space in no time and limit number of containers. Is that an intent?
This should be reduced in RC so that you can spin up about 4096 networks.
As to the reason? It’s the docker default…
I could be wrong here but I’m not sure why it would even matter in most cases. It’s my understanding that this address space isn’t routed anywhere and is just used for L2 communication inside of the container. Overlapping networks wouldn’t exist if they are different broadcast domains, so reusing the same IP space shouldn’t matter.
The only problem would be if you were using 172.16.x.x/12 in your routing table on your LAN.
Yeah, but that’s not the way docker does it. And actually it may matter in the kernel.
Anyway, each docker composition by default makes a unique network from that range. And when it fails it fails to start.
Net effect is a limit of approximately 30 apps in beta.
And 4096 in later versions.