We've seen intermittent integration test failures both locally and in CI (running examples/cpem-add-on) due to TCP timeouts in the addons module. I suspect this is happening because the CPEM example doesn't have a way to ensure that the addons module doesn't start running until after the cluster is ready.
This updates the cluster module to output the IDs of the null_resources that are used to wait for cloud-init and to copy the kubeconfig from the cluster; the idea is that these should roughly be indicators of cluster readiness. The examples/cpem-add-on module has been updated to show how to use these outputs to delay execution of terraform modules that require a running cluster.
We've seen intermittent integration test failures both locally and in CI (running
examples/cpem-add-on) due to TCP timeouts in the addons module. I suspect this is happening because the CPEM example doesn't have a way to ensure that the addons module doesn't start running until after the cluster is ready.This updates the cluster module to output the IDs of the
null_resources that are used to wait forcloud-initand to copy the kubeconfig from the cluster; the idea is that these should roughly be indicators of cluster readiness. Theexamples/cpem-add-onmodule has been updated to show how to use these outputs to delay execution of terraform modules that require a running cluster.