While more than likely never to come up in a real-world situation,
during renewal integration testing a value of 120 seconds has
proven to be too high (the old challenge record has not expired
by the time the new one is created).
Route 53 allows multiple zones with the same name to co-exist in an
account. The most common use case for this is a split-view DNS with one
private and one public zone for the same domain name. This patch makes
sure we don’t ever provision the authorization record to the private
zone.
The other case where a user has multiple public zones with the same
name is not covered here since this would require a bigger change in
code in order to determine which of the zones is active from the
viewpoint of the internet. Also this is probably an edge use case that
can be addressed once it comes up in the issues.
Added a slice of NS to be used when retrying queries. Also used with FindZoneByFqdn()
Adjusted 2 error messages given to better differentiate the returned error string
Fully backwards compatible in terms of credential mechanisms
(environment variables, shared credentials file, EC2 metadata). If a
custom AWS IAM policy is in use it needs to be updated with permissions
for the route53:ListHostedZonesByName action.
Different DNS providers were handling credentials in different ways.
Some were reading credential environment variables in cli_handlers.go
and then passing them into the NewDNSProvider function, while others
were reading the environment variables within their NewDNSProvider
functions.
This change replaces each DNS challenge's NewDNSProvider function with
two new functions: (1) a NewDNSProvider function that takes no
parameters and uses the environment to read credentials, and (2) a
NewDNSProviderCredentials that takes credentials as parameters.
