Long long run MAIN_SHDRUN/ACT_UPG with all component Update

1387739 - Out of memory errors during shadow system operation

 

 

Long long run MAIN_SHDRUN/ACT_UPG with all component Update  sum,ehp ETC

 

 

Symptom

Out of memory error in ACT_UPG phase during EHP installation using SUM.
Dump EXSQL_DSQL_NOROLL in RUN_RADCUCNT_ALL SUM phase.

Other Terms

SUM ACT_UPG RDDMASGL TSV_TNEW_PAGE_ALLOC_FAILED SYSTEM_NO_ROLL TSV_LIN_ALLOC_FAILED RUN_RADCUCNT_ALL EXSQL_DSQL_NOROLL TSV_TNEW_OCCURS_NO_ROLL_MEMORY

Reason and Prerequisites

Mass-Activation program of the ABAP dictionary can run out of memory depending of the kind and number of packages included in the EHP installation.

Solution

Generally the solution is to change the memory parameters of the shadow system so that the dictionary activation program can use more physical memory.

Explanation of some terms used below:

• DIR_PUT can be found using the transaction AL11.

• Depending on the version of the update tool you are using following substitutions might be required:

           SUM -> SAPEHPI
           SAPup -> SAPehpi

Changing memory parameters of the shadow system

The relevant profile parameters of the shadow system and their values recommended for fixing the problem are:

abap/heap_area_total = 10000000000
abap/heap_area_nondia = 10000000000
ztta/roll_area = 20000000

To change the parameter you need to process as follows:

1. Stop the shadow system.

2. Add/change the mentioned parameters in the instance profile of the shadow  system. To find out the right file name and path of the shadow instance profile look in the log file <DIR_PUT>/log/SAPup.log for

              # SHD PROFILE PATH
              # SHD PROFILE NAME

3. Start shadow system.

4. Repeat the ACT phase.

Before changing this parameters make sure you have read the general memory notes 146289, 88416 (MS Windows), 941735 (64-bit Linux).
Preventive setting of the parameters
In some cases you might expect the described error to arise: for example, if the error already occurred during SUM execution on the test or development system and you are now about to update your productive system using the same EHP and SP stack.
If you are absolutely sure that the described error will arise in your planed SUM application you can set the memory parameters for the shadow system preventively in the phase START_SHDI_FIRST:

1. Execute the command

              <DIR_PUT> /bin/SAPup stop START_SHDI_FIRST
              You should execute this command after the update was started but before the phase START_SHDI_FIRST was processed.

2. When the EHPI/SUM stops in START_SHDI_FIRST change the profile parameters as described in the previous section.

              You don't need to start/stop shadow system!

3. Continue with the SUM.

It is not recommended to change the memory parameters of the shadow system in advance as described here if you are not sure that the error will occur! The changed settings may significantly reduce the performance of your productive system during the preprocessing phase of the SUM execution.