Stopping CRS on node0... Stopping CRS on node1... Stopping CRS on node2... Applying patch to Oracle home... Patch 34567890 applied to node0 (1 of 3) Starting CRS on node0...
opatchauto apply /u01/storage/patch_4100 -nonrolling
The tool paused for three seconds—long enough for him to second-guess everything—then spat back:
But the cluster was live. Four thousand active sessions. Three replicas of the order-processing database. If he ran the patch in , nodes would update one by one—seamless, safe, standard. opatchauto-72030 execute in non-rolling mode
He closed the terminal, finished his cold coffee, and wrote the post-mortem email subject line before sleep took him:
OPatchAuto succeeded in non-rolling mode.
Twenty-two minutes later, node0 was back. Then node1. Then node2. Services re-registered. Connections trickled back. Stopping CRS on node0
He pulled up the change request dashboard. His eyes skimmed over the numbers: active transactions, replication lag, customer SLAs. If he did this now, the order system would vanish for at least forty-five minutes. The on-call manager would scream. The VP of Engineering would ask why he hadn’t scheduled a maintenance window.
Log Entry: opatchauto-72030 execute in non-rolling mode Time: 02:13:47 UTC Host: dg-cr1-node0 User: oracle
“I know what it means.”
opatchauto-72030: Prerequisite check failed — rolling mode not supported for this patch on RAC with custom service affinities.
The entire cluster would go down. All nodes. At once. Patch applied to the Oracle home while the databases were offline. Then a full restart. It was the nuclear option—sledgehammer surgery.
