OCP 3.11 non-default SDN setting "OVS Multitenant" causing problems in deployments!

[abudavis]

Describe the bug The OVS Multitenant flag caused our pre-production enterprise grade deployment of our application to fail as it had multiple components deployed across namespaces which were unable to communicate with each other. As a workaround, we had to explicitly allow this communication. We were not aware the script "https://github.com/microsoft/openshift-container-platform/blob/master/scripts/deployOpenShift.sh" changes the OCP 3.11 default SDN setting (which is "openshift-ovs-subnet") to "openshift-ovs-multitenant"

To Reproduce Use scripts as is "https://github.com/microsoft/openshift-container-platform/blob/master/scripts/deployOpenShift.sh" and have an application deployed across namespaces wherein pods of one namespace(s) talks to pods of the other namespace(s)

Expected behavior https://github.com/microsoft/openshift-container-platform/blob/master/scripts/deployOpenShift.sh should either not have the line "os_sdn_network_plugin_name='redhat/openshift-ovs-multitenant'" OR should consider to set to use the OCP 3.11 SDN default "os_sdn_network_plugin_name='redhat/openshift-ovs-subnet'" OR have it as a field in the parameters JSON file

Template Information (please complete the following information):

• OCP : 3.11.153
• OS: RHEL LVM 7.7
• Branch: master

[uselessidbr]

That’s the default behaviour of multitenant SDN.

You can allow projects to communicate using oc adm pod-network

Allowing global or between specific projects