These are known bugs for Oracle Grid Infrastructure in Oracle Database 12c Release 2 (12.2).
When upgrading an Oracle Grid Infrastructure and, if the Oracle Cluster Ready Services (CRS) stack is down before launching the Oracle Grid Infrastructure installer, the following error is returned:
[INS-40418] The installer has detected that Oracle Clusterware is not running on local node
If you bring the CRS stack up and continue with the upgrade, it will fail with the following error:
[INS-20802] Creating Container Database for Oracle Grid Infrastructure Management Repository failed
INS-40418 error is returned, terminate the Oracle Grid Infrastructure installer, bring the CRS stack up, and then restart the Oracle Grid Infrastructure installer.
During an install session, CVU reports a failure in the soft stack shell limits check and the user requests for fix-up and check again action.
Once the fix-ups are generated and executed by the user on the applicable nodes, then the installer needs to be relaunched to load the fixed soft stack shell limits.
The extended redundancy disk group option is not supported in this release of Oracle Grid Infrastructure 12c Release 2 (12.2).
Sometimes starting an Oracle Grid Infrastructure stack fails with the following error stack when there are reader farm services on a pluggable database (PDB):
CRS-5017: The resource action "<rfservice on pdb> start" encountered the following error: ORA-12963: A read-only instance cannot be the first instance to open a database
Ensure that PDB is started on a Hub Node and then restart the service reported as part of the
CRS-5017 error message.
This issue affects rolling upgrade from Oracle Grid Infrastructure 12c Release 1 (12.1) to Oracle Grid Infrastructure 12c Release 2 (12.2) of Oracle Clusterware standard Cluster with Oracle ASM. A node running Oracle Grid Infrastructure 12c Release 1 (12.1) cannot join the cluster after the first node has been upgraded to Oracle Grid Infrastructure 12c Release 2 (12.2).
The nodes running Oracle Grid Infrastructure 12c Release 1 (12.1) that are in
ONLINE status continue to be members of the cluster.
Upgrade Oracle Grid Infrastructure from 12c Release 1 (12.1) to 12c Release 2 (12.2) on the failed node.
It is not recommended to unmount the Oracle Grid Infrastructure file system when installing, upgrading, or patching the grid. However, if this is attempted, the unmount is likely to fail because logging activity may be in progress.
Stop the Oracle Grid Infrastructure stack, unload the Oracle ACFS drivers using the
acfs_load stop command, and then unmount the Oracle Grid Infrastructure file system.