SAP Note 28025 - Shared pool 10: (default size assumed)

Symptom:

Error messages are displayed in the trace file, for example, dev_disp:
*** ERROR => Profile configuration error detected [shmux.c 1962]
*** ERROR => Shared Pool 10: ipc/shm_psize_10 = 2000000 (too small)
*** ERROR => Shared Pool 10: (smaller than default 12000000) [shmux.c
*** ERROR => Shared Pool 10: (default size assumed 12000000) [shmux.c

Other terms

sappfpar, shared memory pool

Reason and Prerequisites

A profile parameter (ipc/shm_psize_10 or ipc/shm_psize_40) for setting the size of a shared memory pool has a value that is lower than the default value. The system uses the default value instead of the specified value.
The system can run without restrictions. No follow-on error are expected.

Solution

The value of the profile parameter (ipc/shm_psize_10 or ipc/shm_psize_40) should not be set lower than the default value and the default disposition should be used for the pools.
If the system creates a pool that is much larger then required, this results in the loss of swapspace only, not main memory. We recommend that you reserve at least 2 MB in the pool.
Deviate from the default disposition of the shared memories only when absolutely necessary. Adhere to the suggested default disposition from the
profile check program sappfpar.
Call: sappfpar check pf=
In some installations, it has been observed that many R/3 shared memories are taken from pools to avoid the dependency of the pool sizes on the size of the contained R/3 shared memories, for example, using parameters such as:
ipc/shm_psize_ = 0.
We strongly advise against this. Reasons:

1. When the R/3 is stopped, certain shared memories are not cleaned up if they are outside of the pool. Problems then occur the next time the R/3 is started. Shared memories with numbers {12, 13, 15, 16, 46, 49} must always be in the pool.

2. Performance problems occur on some hardware platforms (particularly HP) because the higher number of shared memories increases the possibility of protection faults occurring during access. (The HP-MMU contains only a limited number (for example, 6) of protection cache registers for shared memories. These may be set during access using a protection fault and may increase the access time by a factor of 10 ... 40.)