|
Swap Space Bottleneck During
SAP Operation
Symptom Diverse error messages that refer to memory bottlenecks, or, more precisely, a swap space bottleneck. For example, System Panic; cmemreserve: reservation overrun; ENOMEM, Not enough core; ENOSPC, No space left on device; SIGDANGER (under AIX only). In the SAP System, the SAP system log message "no memory of class perm" is output. This problem can occur with SAP processes and external processes as well. The relevant process cannot allocate anymore heap memory. This can lead to the database operation being ended or SAP work processes being stopped. Error messages in the SAP System appear in the developer trace files dev_disp, dev_w<n> in the system log and in dumps. The following messages may appear: TSV_TNEW_;._NO_ROLL_MEMORY; NO_MEM; NO_MEMORY; RESIZE_EM_ALLOC_ERROR, Storage class PERM. This problem mostly occurs if background jobs are active with large amounts of data. Possible Causes There is no more swap space available The highest value for one of the SAP profile parameters that limit the swap space usage has been exceeded. In this case, the following parameters are relevant: abap/heap_area_dia, abap/heap_area_nondia and abap/heap_area_total. The operating system limits for the process size (segment size) have been exceeded. For example, no more that 256 MB can be assigned under AIX and no more than 384 under SINIX. Corrective action Increase the swap space or decrease the SAP Extended Memory. If the swap space already corresponds to the SAP recommendations, 1.5 to 2 GB may still be required. There should always be several hundred MB of free swap space available. You can reduce the maximum swap space need of the SAP System by reducing the value of parameter abap_heap_area_total. You can also distribute the load onto several application servers and increase the roll area. Check if the maximum process size (operating system parameters) is sufficient. ---
Have been installing ECC 6.0 SR3 on azure with linux and
MAxDB. The installation at the phase 'Start Instance' failed to start.
In dev_w0 and dev_disp found about some error as below.
Answer:
Run sappfpar against the instance profile and fix any
errors present in the output.
If the issue happens again then supply the contents of the instance profile - /usr/sap/SD8/SYS/profile/SD8_DVEBMGS00_smartdocs You should also make sure that the SUSE Linux kernel parameters
are set correctly for SAP systems.
I found SAP Note 1597355 which discusses the possible
size of a LINUX OS swap file when SAP is installed. When I reviewed this
file, it appears to be dated from July of 2011. My question(s) are these:
My situation is I'll have RHEL 6 VM's of 64, 96, & 128 GB RAM. If I went by what's in SAP Note 1597355 I'd need to reserve an OS Swap resource of ~100GB for each VM. If I followed Oracle, I'd only allocate 16GB of Swap per VM (a savings of 84GB per VM which is not trivial over 10's or hundreds of VM's). Any advice on this topic would be great. Answer:
It is recommended to configure a small swap space in order to avoid performance regressions at times of high memory utilization on operating system side. Instead it is usually better if activities are terminated with "out of memory" errors. This makes sure that the overall system is still usable and only certain requests are terminated. A good value for the swap space is 2 GB (see e.g. SAP Note 1944799 for SLES environments). |
|
Get help regarding your Basis problems
SAP Basis Admin Books
SAP Basis Tips
Administration In SAP - Sapgui, Unix, SAP ITS, Router, Client Copy and IDES Main Index
All the site contents are Copyright © www.erpgreat.com
and the content authors. All rights reserved.
|