Closed corymhall closed 1 month ago
I'm quite sure we're getting hit by eventual consistency here.
Right now we're doing the following in code:
Depending on how the service implements reads (IIRC almost all are eventually consistent), we're going to run into issues every now and then. My guess is that Networking related resources and IAM related resources are affected the most by this as they have the highest propagation delay IIRC.
What we should do here is add a retry with backoff mechanism in the case the CC operation was successful but the subsequent Read call returns a 404.
Just realized that we already had an issue for this https://github.com/pulumi/pulumi-aws-native/issues/1186
Cannot close issue:
resolution/
Please fix these problems and try again.
What happened
The
AWS::EC2::SubnetRouteTableAssociation
creation seems to be flaky. When a resource is created a subsequentGetResource
call is made to populate the attributes of the resource. Sometimes when creating a SubnetRouteTableAssociation it will fail on theGetResource
call.On a subsequent
up
the error goes away and the deployment completes successfully.Example
This can be seen sometimes in the
pulumi-cdk
tests which create a Vpc.Example tests