Friday Aug 22, 2014

Automatic Maintenance Jobs in every PDB?
New SPM Evolve Advisor Task in Oracle

A customer checking out our slides from the OTN Tour in August 2014 asked me a finicky question the other day:

"According to the documentation the Automatic SQL Tuning Advisor maintenance task gets executed only within the CDB$ROOT, but not within each PDB - but the slides are not clear here. So what is the truth?"

Ok, that's good question. In my understanding all tasks will get executed within each PDB - that's why we recommend (based on experience) to break up the default maintenance windows when using Oracle Multitenant. Otherwise all PDBs will have the same maintenance windows, and guess what will happen when 25 PDBs start gathering object statistics at the same time ...

The documentation indeed says:

Automatic SQL Tuning Advisor data is stored in the root. It might have results about SQL statements executed in a PDB that were analyzed by the advisor, but these results are not included if the PDB is unplugged. A common user whose current container is the root can run SQL Tuning Advisor manually for SQL statements from any PDB. When a statement is tuned, it is tuned in any container that runs the statement.

This sounds reasonable. But when we have a look into our PDBs or into the CDB_AUTOTASK_CLIENT view the result is different from what the doc says. In my environment I did create just two fresh empty PDBs (CON_ID 3 and 4):

SQL> select client_name, status, con_id from cdb_autotask_client;

CLIENT_NAME                           STATUS         CON_ID
------------------------------------- ---------- ----------
auto optimizer stats collection       ENABLED             1
sql tuning advisor                    ENABLED             1
auto space advisor                    ENABLED             1
auto optimizer stats collection       ENABLED             4
sql tuning advisor                    ENABLED             4
auto space advisor                    ENABLED             4
auto optimizer stats collection       ENABLED             3
sql tuning advisor                    ENABLED             3
auto space advisor                    ENABLED             3

9 rows selected.

I haven't verified the reason why this is different from the docs but it may have been related to one change in Oracle Database The new SPM Evolve Advisor Task ( SYS_AUTO_SPM_EVOLVE_TASK) for automatic plan evolution for SQL Plan Management. This new task doesn't appear as a stand-alone job (client) in the maintenance window but runs as a sub-entity of the Automatic SQL Tuning Advisor task. And (I'm just guessing) this may be one of the reasons why every PDB will have to have its own Automatic SQL Tuning Advisor task 

Here you'll find more information about how to enable, disable and configure the new Oracle SPM Evolve Advisor Task:


Grid Infrastructure Management Repository (GIMR)
database now mandatory in Oracle GI

During the installation of Oracle Grid Infrastructure you've had the following option to choose YES/NO to install the Grid Infrastructure Management Repository (GIMR) database MGMTDB:

With Oracle Grid Infrastructure this choice has become obsolete and the above screen does not appear anymore. The GIMR database has become mandatory

What gets stored in the GIMR?

See the changes in Oracle Clusterware here:

  • Automatic Installation of Grid Infrastructure Management Repository

    The Grid Infrastructure Management Repository is automatically installed with Oracle Grid Infrastructure 12crelease 1 ( The Grid Infrastructure Management Repository enables such features as Cluster Health Monitor, Oracle Database QoS Management, and Rapid Home Provisioning, and provides a historical metric repository that simplifies viewing of past performance and diagnosis of issues. This capability is fully integrated into Oracle Enterprise Manager Cloud Control for seamless management.

Furthermore what the doc doesn't say explicitly:

  • The -MGMTDB has now become a single-tenant deployment having a CDB with one PDB
    • This will allow the use of a Utility Cluster that can hold the CDB for a collection of GIMR PDBs
  • When you've had already an Oracle GIMR this database will be destroyed and recreated
    • Preserving the CHM/OS data can be acchieved with OCULMON to dump it out into node view
  • The data files associated with it will be created within the same disk group as OCR or VOTING
    •  The OUI will get the disk groups for OCR and Voting and chooses the first one - which usually is the first OCR. This may lead to serious space issues. It is tracked internally as Bug:19661882  In a future release there may be an option offered to put in into a separate disk group.
      Workaround would be to move the affected OCR to another disk group (use ocrconfig command for it) - see MOS Note:1589394.1
  • Some important MOS Notes:
    • MOS Note 1568402.1
      FAQ: 12c Grid Infrastructure Management Repository, states there's no supported procedure to enable Management Database once the GI stack is configured
    • MOS Note: 1921105.1
      Managing the Cluster Health Monitor Repository (incl how to resize)  
    • MOS Note 1589394.1
      How to Move GI Management Repository to Different Shared Storage
      (shows how to delete and recreate the MGMTDB)
    • MOS Note 1631336.1
      Cannot delete Management Database (MGMTDB) in 12.1
    • MOS Note 1945558.1
      _mgmtdb Service Registered with All Local Listeners in a Grid Infrastructure Environment
  • Average growth size per day per node is roughly 650-750 MB. E.g. a 4 node cluster would lead at the default retention of 3 days to an approximate size of  5.9-6.8 GB
  • Change the retention
    $CRS_HOME/bin/oclumon manage -repos changeretentiontime 260000
  • Change the retention
    $CRS_HOME/bin/oclumon manage -repos checkretentiontime 260000


Markus Michalewicz, our Director of Product Management, Oracle Real Application Clusters (RAC), has published a very interesting and helpful insight article about GIMR on July 30, 2015. Read it here:


PS: Kudos to Sebastian Solbach who updated me on the things to add (retention, average growth, OUI choosing the first disk group displayed for the MGMTDB) - cheers!


Mike Dietrich - Oracle Mike Dietrich
Master Product Manager - Database Upgrade & Migrations - Oracle

Based in Germany. Interlink between customers/partners and the Upgrade Development. Running workshops between Arctic and Antartica. Assisting customers in their reference projects onsite and remotely. Connect via:

- -


« August 2014 »
Slides Download Center
Visitors since 17-OCT-2011
White Paper and Docs
Viewlets and Videos
Workshop Map
x Oracle related Tech Blogs
This week on my Rega & Pono
Upgrade Reference Papers