Created on 12-20-2017 01:19 PM - edited 09-16-2022 05:39 AM
We are using Clouder Director 2.6 to deploy customer clusters in AWS.
We have a need to deploy a new CDH cluster within the same AWS VPC subnet of an existing customer cluster. I need to confirm that there is no chance of service traffic in the new cluster interferring with traffic in the existing production cluster. Do CDH services broadcast in order to discover one another on the same subnet? Is there a chance of services in one custer responding to traffic in another?
We will be migrating data to the new cluster and shutting down the other within a week after spinning up the new cluster.
Thanks
Created 12-21-2017 08:19 AM
dturner,
The clusters will be independent.
There is one minor caveat. Director finds the instances in AWS based on their "Cloudera-Director-Id" tag. Cloudera Director uses UUIDs for these so they will be unique across all instances. However, Cloudera Director accepts any String for this ID if you are creating your cluster through the API (i.e., not through Cloudera Director's UI or CLI). In that case, it is incumbent upon you to ensure you don't have any ID conflicts.
Here's an example of how you could get into trouble.
If you GET a ClusterTemplate through the API, change the "name", and POST it back to create a new Cluster. Director may find the wrong instances in AWS. To avoid this, you must also change all the "id" fields for all of the "virtualInstances".
Created 12-21-2017 08:19 AM
dturner,
The clusters will be independent.
There is one minor caveat. Director finds the instances in AWS based on their "Cloudera-Director-Id" tag. Cloudera Director uses UUIDs for these so they will be unique across all instances. However, Cloudera Director accepts any String for this ID if you are creating your cluster through the API (i.e., not through Cloudera Director's UI or CLI). In that case, it is incumbent upon you to ensure you don't have any ID conflicts.
Here's an example of how you could get into trouble.
If you GET a ClusterTemplate through the API, change the "name", and POST it back to create a new Cluster. Director may find the wrong instances in AWS. To avoid this, you must also change all the "id" fields for all of the "virtualInstances".
Created 12-21-2017 08:45 AM
Thank you for confirming this.
David.