Our primary VPC is configured with four subnets. As you can see below, the four subnets span all of the available ip-addresses within the "10.50.?.?" block. Since all of the available addresses are encompassed within that 10.50 prefix, we are unable to create additional subnets.
We would like to create additional subnets in the 10.50.?.? block in (new) different availability zones because they include (for example) newer instance types. We have a similar issue in other regions.
All of the available ipv4 addresses are not actually in use. However, the web UI will not allow us to "shrink" the existing subets. The way this scenario appears to us, we would have to stop all of our instances to free up the subnets, and remove them before creating new, smaller subnets.
fr-usx-lan-a subnet-a42932c1 vpc-1d0f1b78 10.50.0.0 /18 us-west-2a
fr-usx-wan-a subnet-962932f3 vpc-1d0f1b78 10.50.128.0 /18 us-west-2a
fr-usx-wan-b subnet-357d286c vpc-1d0f1b78 10.50.192.0 /18 us-west-2c
fr-usx-lan-b subnet-1d7d2844 vpc-1d0f1b78 10.50.64.0 /18 us-west-2c
We would like to have access to these subnets as well (in the Oregon region):
2b, 2d
If it is unclear from the text above, we would like to keep using the 10.50.?.? prefix for all instances running in that region. We would also like to avoid having to stop and recreate instances that are currently in production.
Currently, I have to admit that I at a bit of a loss of how to proceed here. I'm thinking some kind of temporary VPC with a bridge between them would be viable. Any suggestions are welcome.
Implement IPv6. One /64 per subnet is effectively unlimited addresses. Use a /56 from AWS for 256 /64s.
Unfortunately, I don't have a v4 option, especially not with the restriction of keeping the same range and staying running.
Classic example of a v4 address plan appearing large, until it needs to be further subdivided into new nets. Don't underestimate work spent in counting hosts and defragmenting ranges.