Best practices for OpenShiftSDN to OVNKubernetes CNI Offline migration
Updated
Migration to OVNKubernetes from OpenShiftSDN had been available since OpenShift Container Platform 4.8. This article lists accumulated wisdom from knowledge base solutions and recent (OCP 4.16 timeframe) updates to documentation. Be sure to start with the Offline migration to the OVN-Kubernetes network plugin overview section of the product documentation in the Networking Guide.
Knowledge base resources
| Resource | Description |
|---|---|
| OpenShiftSDN CNI removal in OCP 4.17 | This article explains which releases of OCP with OpenShiftSDN are upgradable and available for new installations. It also has the removal notice for 4.17. |
| Recommended practice to follow before Openshift SDN network plugin migration to OVNKubernetes plugin | A list of pre-checks to perform before migration, with specific guidance on IP address ranges and when and how to use routingViaHost. |
| Clusters that depend on static routes and/or routing policies on the host network need local gateway mode set during migration | If pods require host networking routes in order to reach some destinations, additional actions are required. |
| Unable to add nodes after OpenShift SDN to OVN migration | Be sure to use the latest RHCOS image when adding new nodes. |
| POD's default CIDR overlaps other external services within network | When migrating to OVNKubernetes there are reserved IP address ranges. |
| OpenShift update stuck after migrating to OVN-Kubernetes with Trident CSI | Customer using the NetApp Trident CSI may need to take additional actions when migrating. |
| OpenShift etcd backup fails with 'FIPS mode is enabled, but the required OpenSSL library is not available' | In a FIPS-enabled environment, make sure you are upgrading to a release where this issue has been solved. |
| Network Cluster Operator stuck in Degraded state due to kube-proxy options during SDN to OVN migration | If you have configured kubeProxyConfig you may need to take additional actions required to migrate |
| An externalIP of Service not reachable from other namespaces within the same cluster after OVN migration | When using externalIP's with externalTrafficPolicy=Local there may be additional actions required to migrated. |
| OVN-InterConnection/OVN Node to pod communication not working in Openshift 4.14 | Cloned disk have been shown to be an issue after an upgrade |
| When doing offline SDN migration, setting the parameter "spec.migration.features.egressIP" to "false" to disable automatic migration of egressIP configuration doesn't work | Updating "egressIP":false in CNO does not prevent egressIP migration |
| Does Red Hat support automation steps/script for migration of SDN to OVN? | Red Hat does not recommend scripting CNI migrations |
| Openshift onboard applications performing reserve dns query may face performance issue after migrating OCP cluster from SDN to OVN | Openshift onboard applications performing reverse DNS is observing latency after migration to OVN. |
| Multitenant isolated ExternalIP is not reachable from other namespaces after OVN migration | When using NetworkPolicies to emulate multitenant isolated namespaces, Services using ExternalIP are not reachable from another namespaces after OVN migration. |
Recent documentation improvements
The following is a list of improvements that have either been made or have been requested to the migration chapter of the networking guide.
Potential bugs to be aware of
| Resource | Description |
|---|---|
| This content is not included.ExternalIP is not reachable from other namespaces after OVN migration | Different behavior between SDN and OVN-K CNI's detected for reaching externalip from a pod in the cluster |
| This content is not included.Issue accessing Service NodePort/ExternalIP with externalTrafficPolicy: Local on multitenant isolated project | Different behavior detected for a pod behind a service with "externalTrafficPolicy: Local." |
| This content is not included.ClusterNetwork not working at first start when multiple subnets are configured | Clusters with two different clusterNetworks and different hostPrefixes will experience some traffic failing between the two clusterNetworks. |
SBR
Category
Components
Article Type