OCP 4 UPI on VMware Cluster Networks

Latest response

I wish to deploy multiple UPI clusters on VMware within the same subnet. For AWS installations the guide indicates that you can define the cluster network range using the install-config.yaml

https://docs.openshift.com/container-platform/4.1/installing/installing_aws/installing-aws-network-customizations.html

Is the same possible with UPI? I assume this would needed if in the same subnet?

Responses

I have several. The node ip is supplied by vmware(from a prior conversation https://access.redhat.com/discussions/4547551 ). The clusternetwork and servicenetwork are inside the cluster and not referenced externally. My install-config.yaml files are identical for the clusters with the only exception being the metadata name (aka cluster "subdomain"). I have an "os" (sandbox) and "on" (nonprod) clusters. You must put quotes around a domain using the name "on", "y", or "yes" otherwise it will be resolved as a Boolean and the resulting domain name will be True.you.org instead of on.you.org. Hope that helps!?

# sandbox cluster yaml

$ cat install-config.yaml.backup.os
apiVersion: v1
baseDomain: you.org
compute:
- hyperthreading: Enabled
  name: worker
  replicas: 0
controlPlane:
  hyperthreading: Enabled
  name: master
  replicas: 3
metadata:
  name: "os"
networking:
  clusterNetwork:
  - cidr: 10.129.0.0/16
    hostPrefix: 24
  networkType: OpenShiftSDN
  serviceNetwork:
  - 10.128.0.0/16
platform:
  vsphere:
    vcenter: your.vcenter.server
    username: vmware_openshift_account
    password: password
    datacenter: "datacenter"
    defaultDatastore: "/datacenter/path/datastore/big_datastore"
pullSecret: <The pull secret that you obtained from the OpenShift Infrastructure Providers page.>
sshKey: '<copy and paste .ssh/id_rsa here>'


# nonprod cluster yaml

$ cat install-config.yaml.backup.on
apiVersion: v1
baseDomain: you.org
compute:
- hyperthreading: Enabled
  name: worker
  replicas: 0
controlPlane:
  hyperthreading: Enabled
  name: master
  replicas: 3
metadata:
  name: "on"
networking:
  clusterNetwork:
  - cidr: 10.129.0.0/16
    hostPrefix: 24
  networkType: OpenShiftSDN
  serviceNetwork:
  - 10.128.0.0/16
platform:
  vsphere:
    vcenter: your.vcenter.server
    username: vmware_openshift_account
    password: password
    datacenter: "datacenter"
    defaultDatastore: "/datacenter/path/datastore/big_datastore"
pullSecret: <The pull secret that you obtained from the OpenShift Infrastructure Providers page.>
sshKey: '<copy and paste .ssh/id_rsa here>'

Thanks Paul! So 10.129.0.0/16 is not an overlay network? In the AWS guide it states:

"A block of IP addresses from which Pod IP addresses are allocated. The OpenShiftSDN network plug-in supports multiple cluster networks. The address blocks for multiple cluster networks must not overlap. Select address pools large enough to fit your anticipated workload."

I believe they are saying the clusternetwork and servicenetwork cannot overlap. Both of these networks are completely internal to a single openshift cluster.

When you say internal to the cluster does that mean this IP(s) is never visible on the physicla/vmware network?

Yes, that is correct. Wolfram Richter has a good 6-part description on OpenShift Networking. It's here: http://www.opensourcerers.org/author/wrichter/