Table of Contents
Status of a failed 12.2 GIMR startup
- You’re starting a 12.2 RAC Database Installation either as
- Standalone RAC cluster or
- Domain Service Cluster
- Your memory Capacity Planning looks like the following
Memory Resources when trying to setup Domain Service cluster with 16 GByte Memory VM DSC System1 [ running GIMR Database ] : 7 GByte VM DSC System2 [ NOT running GIMR Database ] : 6 GByte VM NameServer : 1 GByte Window 7 Host : 2 GByte
- You’re installing the RAC envronment in a Virtualbox env and you have only 16 GByte memory
- The installation works fine until start of the GIMR databae [near at end of the GRID installation process ! ]
- Your Virtualbox WindowsVMs freeze or any the RAC VMs reboots
- Sometimes the start of the GIMR database during the Installation process fails with ORA-3113 or timeout Errors starting a database process like DWB, ..
- Check the related log File in /u01/app/grid/cfgtoollogs/dbca/_mgmtdb/
Typical Error when starting GIMR db during the 12.2 installation
[WARNING] [DBT-11209] Current available physical memory is less than the required physical memory (6,144MB) for creating the database. Registering database with Oracle Grid Infrastructure Copying database files Creating and starting Oracle instance DBCA Operation failed. Look at the log file "/u01/app/grid/cfgtoollogs/dbca/_mgmtdb/_mgmtdb.log" for further details. Creating Container Database for Oracle Grid Infrastructure Management Repository failed. -> /u01/app/grid/cfgtoollogs/dbca/_mgmtdb/_mgmtdb.log Registering database with Oracle Grid Infrastructure Copying database files .... DBCA_PROGRESS : 22% DBCA_PROGRESS : 36% [ 2017-05-04 18:01:38.759 CEST ] Creating and starting Oracle instance DBCA_PROGRESS : 37% [ 2017-05-04 18:07:20.869 CEST ] ORA-03113: end-of-file on communication channel [ 2017-05-04 18:07:55.288 CEST ] ORA-03113: end-of-file on communication channel [ 2017-05-04 18:11:07.145 CEST ] DBCA_PROGRESS : DBCA Operation failed.
Analyzing the problem of a failed GIMR db startup
- During startup the shared memory segment created by GIMR db is about 4 GByte
- If we are short in memory the shmget system call may fail [ shmget(key, SHM_SIZE, 0644 | IPC_CREAT) system ] will kill the installation process by :
- rebooting any of your RAC VM
- freeze your VirtualBox host
- freeze all the the 3 RAC VMs
In any case the installation process gets terminated and you need to cleanup your system and restart the installation process which will fail again untill you don’t add some memory.
Potential Workaround by modifying DomainServicesCluster_GIMR.dbc
After you have extracted the GRID zip file save the DomainServicesCluster_GIMR.dbc and change the following setting [ red : original - blue marks new setting ] Save and Modify DomainServicesCluster_GIMR.dbc [grod@dsctw21 FILES]# diff DomainServicesCluster_GIMR.dbc DomainServicesCluster_GIMR.dbc_ORIG < <initParam name="sga_target" value="800" unit="MB"/> > <initParam name="sga_target" value="4" unit="GB"/> < <initParam name="processes" value="200"/> > <initParam name="processes" value="2000"/> < <initParam name="open_cursors" value="300"/> < <initParam name="pga_aggregate_target" value="400" unit="MB"/> < <initParam name="target_pdbs" value="2"/> > <initParam name="open_cursors" value="600"/> > <initParam name="pga_aggregate_target" value="2" unit="GB"/> > <initParam name="target_pdbs" value="5"/> Copy the modified DomainServicesCluster_GIMR.dbc to its default locaction [grid@dsctw21 grid]$ cp /root/FILES/DomainServicesCluster_GIMR.dbc $GRID_HOME/assistants/dbca/templates [grid@dsctw21 grid]$ ls -l $GRID_HOME/assistants/dbca/templates/Domain* -rw-r--r--. 1 grid oinstall 5737 May 5 14:06 /u01/app/122/grid/assistants/dbca/templates/DomainServicesCluster_GIMR.dbc
- Don’t use this in any Production System – This is for testing ONLY !
-> Now start the installation process by invoking gridSetup.sh in $GRID_HOME !