terraform-aws-openshift
terraform-aws-openshift copied to clipboard
AWS / Kubernetes: Internal DNS is not supported
Note
This may be fixed with the latest version (3.9 at the time of writing) but needs to be tested.
Details
When we use the AWS Cloud Provider (which is required for Persistent Volumes (see #33)), we lose the ability to name our nodes, e.g:
[masters]
master.openshift.local openshift_hostname=master.openshift.local
# host group for etcd
[etcd]
master.openshift.local openshift_hostname=master.openshift.local
# host group for nodes, includes region info
[nodes]
master.openshift.local openshift_hostname=master.openshift.local openshift_node_labels="{'region': 'infra', 'zone': 'default'}" openshift_schedulable=true
node1.openshift.local openshift_hostname=node1.openshift.local openshift_node_labels="{'region': 'primary', 'zone': 'east'}"
node2.openshift.local openshift_hostname=node2.openshift.local openshift_node_labels="{'region': 'primary', 'zone': 'west'}"
Becomes:
[masters]
ip-10-0-1-31.ec2.internal openshift_hostname=ip-10-0-1-31.ec2.internal
# host group for etcd
[etcd]
ip-10-0-1-31.ec2.internal openshift_hostname=ip-10-0-1-31.ec2.internal
# host group for nodes, includes region info
[nodes]
ip-10-0-1-31.ec2.internal openshift_hostname=ip-10-0-1-31.ec2.internal openshift_node_labels="{'region': 'infra', 'zone': 'default'}" openshift_schedulable=true
ip-10-0-1-91.ec2.internal openshift_hostname=ip-10-0-1-91.ec2.internal openshift_node_labels="{'region': 'primary', 'zone': 'east'}"
ip-10-0-1-91.ec2.internal openshift_hostname=ip-10-0-1-91.ec2.internal openshift_node_labels="{'region': 'primary', 'zone': 'west'}"
This does not cause any functional problems, but is frustrating for users as it makes it hard to identify nodes.
The root cause seems to be:
- https://github.com/kubernetes/kubernetes/issues/11543
The following issue is also related:
- https://github.com/openshift/openshift-ansible/issues/5692
Just a note to let you know this issue still isn't fixed with Openshift 3.9 unfortunately.
Thanks for the heads up @bjwschaap, I'll keep the issue open and cross my fingers for 3.10!
This may never be fixed - https://github.com/openshift/openshift-ansible/issues/9665 suggests that openshift_hostname
has been removed, as private 'vanity names' seemed to cause issues. Investigate later
Does this mean we can't run this deployment on a private subnet using private ip/dns names?