So I just recreated the pfile from spfile and looking at the parameter, this is wut i see: orcl2.cluster_database=TRUE orcl1.cluster_database=TRUE When I checked it from the query: select sid, value from v$spparameter where name = 'cluster_database'; SID -------------------------------------------------------------------------------- VALUE -------------------------------------------------------------------------------- orcl2 TRUE orcl1 TRUE But one more thing. When I refresh the database from prod on node 1, I issue following statement before shutting down the db: alter system set cluster_database=false scope=spfile *sid='orcl1'*; Can this sid= clause in the command be creating problem? On Thu, Sep 18, 2014 at 11:01 AM, Yong Huang <yong321@xxxxxxxxx> wrote: > > I then had to recreate the spfile after editing pfile and > > leaving only following: > > > > *.cluster_database=true > > > > But restarting the database from node 1 via srvctl caused > > issue with instance 2 (orcl2) > > Other people's messages are quite clear. I'm not sure where you got stuck. > But maybe you didn't completely cleanup spfile? You can query v$spparameter > to find out. There should be only one entry for this parameter on a RAC > database: > > SQL> col sid for a10 > SQL> col value for a10 > SQL> select sid, value from v$spparameter where name = 'cluster_database'; > > SID VALUE > ---------- ---------- > * true > > If you see more than one > if their values are the same > reset i.e. remove redundant ones (those where sid!='*') > else > reset the wrong ones (or instances wouldn't start; they wanted to > mount exclusively) > > Some parameters must have the same value across all RAC instances. For > lack of a good term, I call them stubborn parameters ( > http://yong321.freeshell.org/oranotes/StubbornParameters.txt). > Cluster_database is one of them. > > Also, double check the spfile location. Ideally, $ORACLE_HOME/dbs has no > spfileSID.ora, but has initSID.ora with just one line in it: SPFILE='<ASM > location of the actual spfile>'. > >