Attempt to improve vpc_subnet / vpc_net stability

[tremble]

SUMMARY

Over in community.aws we're seeing repeated failures due to API rate limits with the subnet/net changes (mostly waiter related)
Switch to a custom waiter that knows how to cope with the API Rate limits.

ISSUE TYPE

• Feature Pull Request

COMPONENT NAME

ec2_vpc_subnet
ec2_vpc_net

ADDITIONAL INFORMATION

Will push changelog once tests complete

[ansibullbot]

cc @brandond @jillr @s-hertel @wimnat
click here for bot help

[alinabuzachis]

Thank you @tremble for asking me to review.

[tremble]

@alinabuzachis For what it's worth I did dig into what it would take to tune the botocore waiters

In theory we could do so by using botocore.client.BaseClient._get_waiter_config() (client._get_waiter_config) however, as an _ method that leaves us reliant on a non-public interface, which I don't think would be a good idea.

This leaves is with copying the waiter configs we care about into waiters.py as the only sane way to inject retrying on API rate limits.

[alinabuzachis]

@alinabuzachis For what it's worth I did dig into what it would take to tune the botocore waiters

In theory we could do so by using botocore.client.BaseClient._get_waiter_config() (client._get_waiter_config) however, as an _ method that leaves us reliant on a non-public interface, which I don't think would be a good idea.

This leaves is with copying the waiter configs we care about into waiters.py as the only sane way to inject retrying on API rate limits.

@tremble Thank you for explaining. It makes sense to me. I am going to add a waiter even for NatGatewayAvailable since I used the built-in one. https://github.com/ansible-collections/community.aws/pull/445/files#diff-847118175f7fbd015daf612a3319157ca22874f706c01f26348f4f86f67384d9R706-R709

[tremble]

Change approved and Shippable tests passing. Merging