Oracle RAC 26ai New Feature

Joining Inaccessible Nodes After a Forced Cluster Upgrade (Oracle Grid Infrastructure 26ai)

With Oracle Grid Infrastructure 26ai, cluster lifecycle management becomes more flexible, especially when dealing with partially upgraded or temporarily inaccessible nodes.

In earlier releases, if a node was unreachable during a cluster upgrade, the only practical option was often to delete and reconfigure the node. In 26ai, Oracle introduces a cleaner and safer mechanism to rejoin inaccessible nodes after a forced cluster upgrade.

 

Scenario: Forced Cluster Upgrade Completed

You performed a force cluster upgrade, and one or more nodes were inaccessible during the process.

Instead of:

• Deleting the node
• Cleaning OCR references
• Re-adding the node from scratch

You can now rejoin the node directly, provided that:

Oracle Grid Infrastructure 26ai software is already installed on the node.

 

🛠 Procedure: Join an Inaccessible Node

Step 1 – Log in as root

On the node that was inaccessible:

ssh root@inaccessible_node

 

Step 2 – Change to Grid Infrastructure Home

cd $GRID_HOME

Example:

cd /u01/app/26.0.0/grid

 

Step 3 – Run the Join Command

Use the following syntax:

./rootupgrade.sh -join -existingnode upgraded_node

Where:

• upgraded_node = A cluster node that was successfully upgraded
• The script synchronizes cluster metadata and configuration

 

Example:

./rootupgrade.sh -join -existingnode node1

This command:

• Reintegrates the node into the cluster
• Syncs OCR configuration
• Aligns voting disk and cluster registry metadata
• Avoids full node reconfiguration

 

Changing the First Node for Installation or Upgrade

Cluster installation/upgrade designates a first node that initializes cluster configuration.

But what if the first node becomes inaccessible?

During Installation

If root.sh fails on the first node:

Run this on another node:

root.sh -force -first

This forces the new node to assume the role of the first node for installation.

 

During Upgrade

If the first node fails during upgrade:

rootupgrade.sh -force -first

This command:

• Overrides first-node designation
• Continues upgrade process from another node
• Prevents rollback or cluster restart requirement

 

Architecture Impact

In large RAC environments:

• Rolling upgrades are common
• Network partitions can occur
• Temporary node failures are realistic

 

With 26ai:

✔ No need to delete/recreate nodes
✔ Less downtime risk
✔ Better operational continuity
✔ Simplified recovery from partial upgrades

This is particularly valuable in:

• Exadata environments
• Extended clusters
• Multi-site RAC with Data Guard

 

 Important Notes

• The node must already have 26ai Grid binaries installed
• Ensure cluster interconnect and voting disks are reachable
• Verify CRS status after join:

crsctl check cluster -all

• Always validate cluster health post-operation:

olsnodes -n

crsctl stat res -t

 

 Summary

Oracle Grid Infrastructure 26ai significantly improves cluster resilience by allowing:

• Rejoining inaccessible nodes after forced upgrades
• Forcing a new first node during install or upgrade

This eliminates the painful delete-and-readd cycle from previous releases and reduces operational complexity in production RAC environments.