KB483972: Settings that need to be checked before starting the Intelligence Server on Linux

Setting 1: mstr_check_max_semaphore

If you do not enlarge the kernel.sem file, you will receive the following warning upon starting the Intelligence Server:
mstr_check_max_semaphore: WARNING: maximum number of semaphore arrays 1024 is low; please increase to 2048 or greater to use high performance share memory IPC. 
 

Modify and verify the setting

1. Add the following in /etc/sysctl.conf using VI or another plain text editor:
   
   kernel.sem=250 32000 32 2048 
   
   If the kernel.sem line already exists, modfy the existing values to: 250 32000 32 2048
   
2. Switch to the root user to make the settings effective immediately.
3. Run /sbin/sysctl -p in the terminal.
4. Check the SEMMNI value by running cat/proc/sys/kernel/sem in the terminal.

For more details, please see the KB34895: How to set SEMMNI in Rehat Enterprise Linux.
 

Setting 2: vm.max_map_count

Strategy recommends that the Linux kernel setting vm.max_map_count be defined as 5,242,880 bytes. This allows the Intelligence Server to utilize the system memory resources. If a lower value is used, the Intelligence Server may not able to use all available system resources. The lack of system resources can cause some Intelligence Server actions to fail. By increasing the value for this kernel setting, additional system resources will be available for the Intelligence Server to successfully execute actions. 
If the vm.max_map_count is not large enough, no core would be generated if the Intelligence Server crashed during concurrency testing. 
 

Modify and verify the setting

1. Add the following in /etc/sysctl.conf using VI or another plain text editor:
   
   vm.max_map_count = 5242880
   
   If the vm.max_map_count line already exists, modfy the existing values to: 5242880
   
2. Switch to the root user to make the settings effective immediately.
3. Run /sbin/sysctl -p in the terminal.
4. Check the value by running /sbin/sysctl vm.max_map_count or cat/proc/sys/vm/max_map_count in the terminal.

Setting 3: ulimit parameters

You can check the ulimit parameters by running ulimit-a in the terminal. Before starting the Intelligence Server, you need to check the following two parameters:

• If "core file size" is set to 0, no core file will be generated if the Intelligence Server crashed. You can change the value by ulimit -c unlimited in the same terminal that the Intelligence Server is going to be started in. 
• If "open files" is too small, you may receive an error messaging stating there are too many files open. This is because you have reached the limit of opened files. You need to increase the ulimit parameters by running ulimit -n 65534. The suggested number here is 65534, however, you can change the value as long as it is large enough for your test. 

Setting 4: core path 

The default value is the following: 

You can find the core dump file under the installation directory of the Intelligence Server.
 

Modify and verify the setting

• Run the following command in the terminal:
  
  echo "/MSTR/core/core,%e.%p" > /proc/sys/kernel/core_pattern
  
• To make the changes permanent, add the following line to /etc/sysctl.conf:
  kernel.core_pattern = core.%e.%p
  
   
  

Setting 5: disk and memory

Modify and verify the setting

• Check for the available disk space: (df-h). Make sure the disk is not full.
• Check for the memory: (free-m). Make sure the memory has 8GB or more.

Setting 6: availability of the folders used by the Intelligence Server

Some directories used by the Intelligence Server, such as directories to store cube, cache, and inbox message, might be configured to a network path. Make sure they are accessible before starting the Intelligence Server. Otherwise, as expected, it may cause the Intelligence Server to hang.
For more details, please see:

• KB483889: Side effects of cache or cube folders not shared correctly in a MicroStrategy Intelligence Server cluster environment
• KB483342: On Linux, the Intelligence server remains in a “starting” state due to a file lock on the EMC Isilon OneFS network file system