To help with troubleshooting.
What is this signal? What to do to avoid such an issue?
What is this signal?
When the number of free paging-space falls below a threshold the operating system informs all processes by sending the SIGDANGER (signal 33) signal of this condition. The AIX operating system terminates the processes that caused this memory bottleneck, among others the SAP processes. This is what shoud be avoided.
What to do to avoid such an issue?
- First of all check how much memory is available: summarize RAM + Swap. This forms the available virtual memory.
- Check how much memory the SAP system is able to use in a worst case scenario. This must be lower than the available virtual memory.
- ABAP: total, worst case requirement is reported at the bottom of sappfpar check pf=<path_to_profile/instance_profile> output.
- Java: Maximum possible heap usage of every nodes (server processes, dispatcher process, SDM, set by -Xmx parameter)
- Database: buffer + some overhead for database processes. Better you summarize ipcs -m <dbstring> , e.g. ipcs -m <ora> (bytes row).
- Summarize every components configured on the host and add 10-15% of RAM for OS and about 3% for filecaching. This will be the approximately worst case requirement.
- Configure sufficient paging space. Optimistic: 20 GB, realistic: 1-1,5 x RAM. Recommended: realistic. See SAP note 1121904.
- Set AIX Virtual Memory Management settings. See SAP Note 973227.