Joan: The rolling cpu patch on RAC nodes works as follows: 1. Suppose we have 3 instances on 3 nodes. Only the "Rollable CPU" patches can be applied this way. First you shutdown one instance, keeping the other 2 instances open and running. Apply the cpu patch to the oracle binaries of node 1. Use the -local command with the apply command. You have to apply locally. If you forget this, it will mess up the binaries of the other 2 nodes. After the successful application of the cpu patch on node 1 binaries, then start the instance on node 1. DO NOT apply any "Post-Installation" tasks to the database yet. 2. Follow step 1, for the other 2 nodes. 3. After the successful application of the cpu patches to oracle binaries on all the 3 nodes, then, you have to coordinate with the user community for the database shutdown. Note that the database is up and running and the users use the database on 2 nodes, while you are applying the patch to the binaries of the 3rd node. 4. After fixing a date and time for the shutdown, then shutdown the database, (which will stop all the instances), apply the "Post-Installation" tasks to the database, and start the database. Even for the rolling patches, the database must be shutdown at some point. Only the oracle binaries, can be patched in the "rolling" fashion without shutting down the database. If and only if the patch requires "Post-Installation" tasks on the database, then the database must be shutdown for that period to apply the Post_Installation tasks. It will be usually about 15-30 minutes. Thanks. Ram Srinivasan On Mon, Aug 30, 2010 at 1:00 PM, Hsieh, Joan <Joan.Hsieh@xxxxxxxxx> wrote: > Hi Listers, > > > > I know that we can use rolling upgrade method to minimize the down time > within the data guard environment. > > How about the cpupatch? Is it possible to use the same mechanism to roll > the cpupatch without shutdown the db? > > > > Thanks, > > > > Joan > -- Sincerely Ram Srinivasan