It looks like the SAP Bundle Patch 12.1.0.2.170418 – 201705 contains some bugs. At least for me, this was noticed in a Solaris Intel environment on a Solution Manger 7.1 and a PI 7.4 system.
After a week or so since the installation, the database was growing unusually fast. Checking the top growth segments in the Solman database, the object responsible for this was the LOB segment corresponding to the column LAST_TEST in table MES_DB_AGGREGATE. In the productive system this was growing with more than 20GB/per day. But let’s see in case you also encounter such rapid growth of the database how can you track what is causing this and since when?
As a first step I would propose to check with the help of tcode ST04 or DBACOCKPIT the top growth of segments via: Space -> Segments -> Overview. Here you can check the top growing segments per day/week/month:
Also here at the tab Top sizes you can check the objects that occupy the most space.
To see how did the growth of a particular segment developed lately, you can check its history growth by going to Space -> Segments -> Detailed Analysis -> Fill in the segment name -> OK
Here you can have a look on how this segment grew on days/weeks/months:
The quickest workaround was to reorganize the responsible segment, in order not to get an overflow of the database.
The reason for this growth was caused by the abnormal behavior of the Oracle SMCO process that makes some LOB segments to extend indefinitely, bug that comes together with the instalation of the bundle patch. To summarize in a few words SMCO (Space Management Coordinator) performs proactive space allocation and space reclamation and coordinates the following space management tasks: tablespace-level space pre-allocation, securefile lob segment pre-extension, temporary segment space reclamation, securefile lob segment space reclamation. There is a parameter that enables/disables the SMCO background process: _ENABLE_SPACE_PREALLOCATION. By default this feature is turned on, so the value is by default set to 3, that means that space preallocation is active for all auto extensible datafiles and securefile segments. Another possible value is 1 and this means that pace preallocation is done only for the datafiles. If you want to quickly get rid of this problem, a temporary solution would be to completely disable this feature and set the parameter to 0:
ALTER SYSTEM SET “_ENABLE_SPACE_PREALLOCATION” = 0;
The good news is that the parameter is dynamic, so no restart is required in order for this to get activated.
A strange behavior of LOB segments growing too large I experienced also on a PI system 7.4 for the LOB segments corresponding to tables BC_MSG and SXMSCLUP.
For more information about LOB segments, you can check out my other blog post: Introduction to LOBs
*** Update: SAP documented in the meantime this behavior of LOB segments in the OSS note: 2538588 – 12c: Extensive LOB segment growth after 12.1.0.2 SBP 201705 or 201708 has been applied and also the fix of the bug can be found now in the newest SAP bundle patch SBP 201711.
