It's possible that when the number of IPs available gets low in a subnet, many instances can request IPs in the same subnet, which result in many IP allocations failing since there are more IPs requested than are available.
Instead, alleviate this issue by picking subnets at psuedo-random, and making sure we retry different subnets on different attempts.
It's possible that when the number of IPs available gets low in a subnet, many instances can request IPs in the same subnet, which result in many IP allocations failing since there are more IPs requested than are available.
Instead, alleviate this issue by picking subnets at psuedo-random, and making sure we retry different subnets on different attempts.