add RHOCS support

[bengland2]

goal is to be able to test OCP 4.3 at scale using the openshift-scale repos using RHOCS storage. There are two things that are needed:

• add RHOCS to deployment and bring it up
• deploy workloads to run using it

Joe and Ravi explained how to do the first task. We can define ansible vars needed by RHOCS in the install-on-aws.yaml file here. Then we can add deployment steps for OCS in the post-install playbook here. OCS automatically deploys storage classes for Ceph RBD and Cephfs, so OpenShift workload generators can reference these.

To deploy OCS you need to create storage server instances of the appropriate number and type, and label them as storage hosts. Code would look something roughly like this:

<generate install-config.yaml  with right server instance types and parameters for AWS>
openshift-install create cluster --dir=.
export KUBECONFIG=`pwd`/auth/kubeconfig
# now deploy OCS on that cluster
git clone https://github.com/openshift/ocs-operator
cd ocs-operator
# fetch appropriate release
oc create -f deploy/deploy-with-olm.yaml
wait for an ocs-operator pod to appear in the openshift-storage namespace
for n in $(oc get nodes | awk '/worker/{print $1}') ; do 
  oc label nodes $n cluster.ocs.openshift.io/openshift-storage='' || exit 1
done
oc create -f deploy/crds/ocs_v1_storagecluster_cr.yaml

But in ansible style with tasks, vars, etc. I think the workload generator hosts have to be created in the same openshift-install run as the RHOCS servers, but with different instance type parameters. How does that work?

@ekuric what do you think?

[bengland2]

@chaitanyaenr @jtaleric is this reasonably close to what we talked about?