1. Real Application Clusters Administration and Deployment Guide
  2. Pluggable Database Rank

Pluggable Database Rank

The PDB -rank parameter defines relative importance of the PDBs, which are created specifying cardinality, in a database with the RANK management policy.

The Pluggable Database (PDB) rank is a predefined value that you can assign to a PDB to specify workload importance of the PDB. Oracle Clusterware makes several decisions based on the rank of the PDB. By default, Oracle Clusterware has the same workload importance for each PDB. However, with the PDB -rank parameter, you can choose from a set of predefined values to distinguish workload importance of a PDB. The higher rank means the higher PDB workload importance, for example, the PDB rank 5 is the highest rank and 1 is the lowest rank.

The PDB -rank parameter is optional, and it is not set by default. You can configure it using the srvctl modify pdb command. When the -rank parameter is set, Oracle Clusterware gives precedence to the PDBs with ranks to perform the following operations:

  • Determines the startup order of the PDBs in the cluster. Oracle Clusterware attempts to start PDBs with the highest rank before the other PDBs with the lower ranks.
  • Shuts down a cluster database instance, with the RANK management policy, if there are no running PDBs that require that database instance.
  • Decides whether to refuse starting the PDBs or stop running the PDBs when the cluster does not have sufficient resources to meet resource requirements of the PDBs with higher ranks when PDB's resource requirements are set to non-default values. The PDBs with non-default rank and resource requirement values have higher priority than the PDBs with default rank and resource requirement values.

    If a PDB in CDB1 has RANK 3 and a PDB in CDB2 has RANK 2, and if there are only enough resources to start only one CDB, then Oracle Clusterware starts CDB1 by dependency because the PDB in CDB1 has a higher rank. Oracle Clusterware does not start CDB2 because the PDB in CDB2 has a lower rank.

How PDB Rank Works?

If the PDB -rank parameter is defined, then Oracle Clusterware first considers PDB with the highest rank and then considers number of required CPUs while failing over the PDBs. For example, in a three node cluster with four CPUs in each node, and three PDBs named PDB1 with RANK 1 and CPU count 4, PDB2 with RANK 2 and CPU count 4, and PDB3 with RANK 3 and CPU count 4, Oracle Clusterware handles failover in the following order:

  1. The cluster has enough resources to start all PDBs. When the first node fails, PDB2 and PDB3 keep running, but Oracle Clusterware stops PDB1 because it has the lowest rank. If PDB2 and PDB3 were hosted on the failed node, then Oracle Clusterware stops PDB1 on the running nodes and starts PDB2 and PDB3 on these nodes.
  2. When the second node fails, PDB3 keeps running, but Oracle Clusterware stops PDB2 because it has the lowest rank. If PDB3 was hosted on the failed node, then Oracle Clusterware stops PDB2 on the running node and starts PDB3 on that node.
  3. PDBs can run from different CDBs. If you have multiple CDBs and any of these CDBs does not have a running PDB, then Oracle Clusterware shuts down the CDBs that do not have any running PDBs.

The rank of a PDB is set for the entire cluster, not only for the CDB in which you create a PDB.