Review this information to understand rolling upgrade of Oracle Grid Infrastructure.
When you upgrade Oracle Grid Infrastructure, you upgrade the entire cluster. You cannot select or de-select individual nodes for upgrade. Oracle does not support attempting to add additional nodes to a cluster during a rolling upgrade. Oracle recommends that you leave Oracle RAC instances running when upgrading Oracle Clusterware. When you start the
root script on each node, the database instances on that node are shut down and then the
rootupgrade.sh script starts the instances again.
You can use
root user automation to automate running the
rootupgrade.sh script during the upgrade. When you use
root automation, you can divide the nodes into groups, or batches, and start upgrades of these batches. Between batches, you can move services from nodes running the previous release to the upgraded nodes, so that services are not affected by the upgrade. Oracle recommends that you use
root automation, and allow the
rootupgrade.sh script to stop and start instances automatically. You can also continue to run
root scripts manually.
Restrictions for Selecting Nodes for Batch Upgrades
The following restrictions apply when selecting nodes in batches for upgrade:
You can pool nodes in batches for upgrade, up to a maximum of three batches.
The local node, where Oracle Universal Installer (OUI) is running, must be upgraded in batch one.
Hub and Leaf Nodes cannot be upgraded in the same batch.
All Hub Nodes must be upgraded before starting the upgrade of Leaf Nodes.