TSV_TNEW_BLOCKS_NO_ROLL_MEMORY

[ashish_vikas]

Hi friends,

I am analyzing one job failure in our system and need your inputs for finding root cause.

We get dump TSV_TNEW_BLOCKS_NO_ROLL_MEMORY.

My first question is, why this process is not using Extended memory ? (when this job was running, i did not see any Extended Memory utilization by other users on this app.

Dump says :

Memory consumption                                                                            |

|    Roll.... 5719808                                                                              |

|    EM...... 0                                                                                    |

|    Heap.... 16504864                                                                             |

|    Page.... 40960                                                                                |

|    MM Used. 13515728                                                                             |

|    MM Free. 8637304   

In wp 48 trace, i see when termination happened, it was the only one which was using heap memory.

A  *** ERROR => heap memory WP48: 3742 MB [abstor.c     1874]

A  {root-id=0050568B642E1ED4ADB291E13038963E}_{conn-id=8FEB4DE0B58811E48248F9DDA90A1DBF}_534

We have :

abap/heap_area_nondia = abap/heap_area_total = 4000683008

ztta/roll_extension_nondia = ztta/roll_extension = 4000683008

em/initial_size_MB = 16384 MB

Regards,

Ashish

TSV_TNEW_BLOCKS_NO_ROLL_MEMORY.txt.zip

[divyanshu_srivastava3]

Hi Ashish,

For your dump, a memory contention/conflict @ "16.02.2015 04:42:33" was the reason of this dump and no use of EM are as follows.

You have a limit on the amount of heap memory that non-diag process can use.

which is - abap/heap_area_nondia = X gigs

Now that X gigs was not reached because the requested chunk was not given to this process for expanding the internal table.

Possible reasons ?

Either you application server was loaded.

OR The data which is job was processing was huge.

OR the Internal table was asking for block of space which cannot be allocated because the address space at that moment was not collected, thus calloc/malloc failed.

Until your heap is fully consumed, why would a process go into extended memory ?

That is the reason why you see '0' in EM.

Regards,

P.S.  I could not read your dump completely, however, the it looked quite generic to me.

[former_member182657]

Hi Ashish,

why this process is not using Extended memory ?

As per understandings report execution was in batch steps.For batch steps process looks first for

   a. Roll memory up to ztta/roll_area
   b. HEAP memory up to abap/heap_area_nondia for background processes or until HEAP memory exhausted (abap/heap_area_total).
   c. EM memory up to  ztta/roll_extension or until total Extended  memory exhausted (em/initial_size_MB)
      -> DUMP

In your issue report generating dump before entering into EM memory area.For this could you share results after following SAP Note 2092663 - TSV_TNEW_OCCURS_NO_ROLL_MEMORY Dump

In addition could you share parameter value for ztta/roll_first and system logs from SM21 during execution of report ?

Hope this will help you.

Regards,

Gaurav

[former_member185954]

Hello Ashish,

There is also a wiki page, which is quite useful Memory Allocation and Analysis - Application Server Infrastructure - SCN Wiki

Regards,

Siddhesh

[manumohandas82]

Hi

Is the work process getting into PRIV mode .

How much physical RAM do you really  have and How much SWAP space ?

ztta/roll_extension_nondia = ztta/roll_extension = 4000683008 . You can  increase the non DIA to 5/6GB

Also can you post your ST02 screenshot .

Thanks ,

MANU

[ashish_vikas]

Why should I increase it to 5/6 GB when i do not see 4GB is being used ?  In-fact, i see this this is my question and concern, why it is not being used.

Regards,

Ashish

[former_member185954]

Hello Ashish,

Check if someone intentionally or accidentally modified the sequence of allocation using report RSMEMORY

Documentation about the program can be found here.

Regards,

Siddhesh

[manumohandas82]

Hi Ashish ;

How do you say that your extended memory is  Not full . And how do you think we will automatically  know that ?

A  *** ERROR => heap memory WP48: 3742 MB [abstor.c     1874]

A  {root-id=0050568B642E1ED4ADB291E13038963E}_{conn-id=8FEB4DE0B58811E48248F9DDA90A1DBF}_534

The above shows that your Heap memory was used .the error normally happen when you have no extended memory as well

Did your work process get into PRIV mode ? Did it restart ?

What your high water mark on your extended memory ?

Send the ST02 screenshot

Thanks ,

Manu

[Sriram2009]

Hi Ashish

The job which your are mention required more memory resource to complete . For this dump "TSV_TNEW_BLOCKS_NO_ROLL_MEMORY" you have to enable the trace using transaction code ST12 trace, then check the overall usage of ABAP & DB, I think it may be problem in the custom code program(Z -Program), you may require to check the program either its takes index or table scan? because of this dump your system overall performance may degrade.

BR

SS

[ashish_vikas]

Did you see the dump ? There is no Z objects involved.

My first question is, why this process is not using Extended memory.

Regards,

Ashish

[alwina_enns]

Hello Ashish,

the program was running in a batch work process? Please take a look at the note:

941735 - SAP memory management system for 64-bit Linux systems

how the memory is allocated for a non-dia work process on Linux (from dump the OS is Linux?)

Regards,
Alwina

[ashish_vikas]

Linux.

we are running job, so BTC work process.

Regards,

Ashish