I am working in an environment which is a 8 node RAC on 10.2.0.3 with 19 patches . Most of them have been applied in a rolling manner. Tonight I have to apply 4 more patches and all of them will be applied in a rolling manner. The process is simple Shutdown services on node a shutdown abort the instance on node a opatch apply -local -oh /opt/oracle/product/10.2.0/db_1 startup the instance So on and so forth across all 8 nodes. After all 8 nodes are patched check that the checksum of the oracle executable is the same across all nodes. The process will take atmost 2 hours. So I am wondering why it should take 10 hours in your case Thanks Fairlie Roman Podshivalov <roman.podshivalov@xxxxxxxxx> wrote: Hi, Here is what we have done in the past: - backup of Oracle home + oraInventory before the change on each node - opatch apply -local on one node at a time - to minimize impact and application downtime because app can still work on unpatched nodes. Also if you hit any installation errors effect will be local to a single node - hopefully. The real downtime stats when you approach the last node. But it's still minimal because once you shut it down you can immediately bring up rest of already patched nodes, if you don't need to run anything in the database, but if you do - you can bring up just a single node and do it, especially if you need to do it with clustered_database parameter set to false - like it was the case with earlier security patch installations. And for 10g there is rolling patch apply - but I've never personally tried it. --romas On 2/20/08, Anurag Verma <anuragdba@xxxxxxxxx> wrote: Hi All, One questions on patching on RAC database. We will be patching our 4 node production RAC this weekend. The database is in 10.2.0.3.0 and the patches are a couple of one-off streams patches and not any major patches. Only difficulty we found is that long time, it takes for applying each patches. (Total time around 10 hours). I was thinking about various backout options. One is rolling back the applied patches, if something goes wrong. Also, another plan is to take a backup of ORacle home from the production RAC nodes. And if we need to backout, just untar the Oracle home backup tar file to the oracle home directory. Please let me know if you have any suggestions. Thanks, -- Anurag Verma, Database Administrator ERCOT(Electric Reliability Council of Texas), Texas 76574 Fairlie Rego Senior Oracle Consultant http://el-caro.blogspot.com/ M: +61 402 792 405 --------------------------------- Get the name you always wanted with the new y7mail email address..