According to the upgrade guide [0] resource names cannot start with digits.
Currently both routes and VPC CIDR associations start with digits, so this adds prefixes to them so that they are valid resource identifiers in 0.12.
This is a significant change because on its own, terraform will destroy and recreate the route which impact the cluster networking.
To avoid this, existing clusters this will require moving the resources within the terraform state prior to the next `apply`.
```
kops update cluster --target terraform --out ./
terraform state mv aws_route.0-0-0-0--0 aws_route.route-0-0-0-0--0 # repeat for all aws_route resources
terraform plan
terraform apply
```
The exact terraform state command may vary depending on how Kops' terraform output is used.
See the command documentation [1] for more details.
Always run a terraform plan first to ensure the `aws_route` and `aws_vpc_ipv4_cidr_block_association` resources are not getting recreated.
Due to the potential impact, this notice should be very prominant in the Kops release notes
[0] https://www.terraform.io/upgrade-guides/0-12.html
[1] https://www.terraform.io/docs/commands/state/mv.html
I made a mistaken assumption in
dde2100a19 that we only had one subnet
per AZ, but as demonstrated in #5587 this was not the case.
What I was trying to achieve was not to include the cluster name, so
for the case of subnets this commit just uses the subnet name from the
cluster spec, which should be unique and stable. That is hopefully at
least as meaningful.
Thankfully we hadn't released a version with the erroneous naming.
Fix#5587
Peter Rifel