Register Login

SAP HANA DB: Crash of the indexserver due RTT issue

Updated May 19, 2018

The indexserver of the SAP HANA database crashes in the method PersistenceManagerImpl::writeRestartData.

In normal case, the rollback transaction table (rtt) contains only a small set of in-doubt rollbacked transactions..
Currently, long log recovery spanning many crashes/restarts of individual services or many rolled back distributed transactions can cause RTT overflow, since RTT is limited to about 500K entries. This may lead to an indexserver crash followed by a not starting indexserver. This is related to Revision 26 or previous.

To be able to restart the indexserver you will need to define the following two parameters 'default_stack_size_kb' and 'worker_stack_size_kb =4096' in the indexserver.ini file:
indexserver.ini
->threads
  ->default_stack_size_kb =4096
  ->worker_stack_size_kb =4096

Those two parameters needs to be defined in the indexserver.ini file
/usr/sap/<sid>/SYS/global/hdb/custom/config/indexserver.ini

Please copy the following 3 lines in the indexserver.ini file and restart the databse:

[threads]
default_stack_size_kb = 4096
worker_stack_size_kb = 4096

After the restart of the database, please delete those two parameters by using the following statements:

ALTER SYSTEM ALTER CONFIGURATION ('indexserver.ini', 'SYSTEM') UNSET ('threads', 'default_stack_size_kb') WITH RECONFIGURE;
ALTER SYSTEM ALTER CONFIGURATION ('indexserver.ini', 'SYSTEM') UNSET ('threads', 'worker_stack_size_kb') WITH RECONFIGURE;

This bug will be fixed with Revision 27.

Read more on SAP HANA


×