Database nodes may have an old configuration file overriding the number of Huge Pages that are automatically configured. This may result in incorrect Huge Pages amount after resizing a node.
After VM resize, the number of HugePages may stay at the old value.
After downsizings a node may fail to start.
FlashGrid Cluster on AWS/Azure/GCP
FlashGrid Storage Fabric
Clusters that were initially deployed with FlashGrid Storage Fabric versions lower than 20.8.53 (released in September 2020) and later updated with node update package versions older than 22.03.80 (released in May 2022)
With automatic configuration of HugePages enabled on database servers, recent versions (20.8.53 or newer) of FlashGrid Storage Fabric will update settings in the following file that is created during boot:
Older versions of FlashGrid Storage Fabric created a different file:
Node update package versions older than 22.03.80 did not remove the second file. As a result, both files may exist supplying conflicting settings.
flashgrid-cluster command to determine which version of FlashGrid Storage Fabric you have.
For version 20.08.42 or older: Your system currently is not affected. DO NOT follow the steps below. Applying node update package version 22.03.80 or newer will prevent the issue from happening.
For version 20.08.53 or newer: Your system may be affected. Follow the steps below.
To determine whether your system is affected, check for the existence of the old file on each database node in the cluster:
If the file is not present, your system is not affected and no further action is needed.
If the file is present,
- Delete the file.
- To apply correct HugePage settings, gracefully reboot the node (use
flashgrid-node rebootafter stopping all Oracle services on the node).
- After reboot, run FlashGrid Health Checker to confirm that database memory settings are consistent with the updated HugePages setting.