When you try to use srvctl to start/stop a database you got this error:

> srvctl start instance -d EAI01H -i EAI01H1 
PRCR-1013 : Failed to start resource ora.eai01h.db 
PRCR-1064 : Failed to start resource ora.eai01h.db on node exa101s01 
CRS-5017: The resource action “ora.eai01h.db start” encountered the following error: 
ORA-01017: invalid username/password; logon denied 
. For details refer to “(:CLSN00107:)” in “/app/base/grid/diag/crs/exa101s01/crs/trace/crsd_oraagent_oracle.trc”. 

CRS-2674: Start of ‘ora.eai01h.db’ on ‘exa101s01’ failed 

root cause:

We’ve deleted an OS group which was accidentally created in the past. During installation time of GI all the OS groups will be written to a config file $GI_HOME/rdbms/lib/config.o

When using srvctl to start/stop a DB srvctl will check this config file and check if oracle binary file has the proper ownership as per config file or not.

If the executable doesn’t have the proper ownership srvctl will use setasmgidwrap to change it as per config file.

solution:

To solve this issue change the file $GI_HOME/rdbms/lib/config.c and relink Oracle binary. How to relink the GI_HOME is described within MoS note 1508027.1

Note: read this note carefully and think about every step before you execute it. This note contains several errors!

To relink the GI_HOME oracle binary you’ve to stop all databases and the crs stack!

workaround:

When you’re not able to relink the binary you can change to oracle binary in the GI_HOME manually but to prevent setasmgidwrap from changing it back to a wrong group you’ve to rename the setasmgidwrap executable within $GI_HOME/bin directory.