On our previous system before exadata, if we had said that we would upgrade the sun machine, the storage boxes, the network switches and also the oracle software at the same time, the change management team would have only laughed at us. But this is exactly what we have done last Friday.
We have upgraded 3 network switches, 8 compute nodes, 14 storage nodes and also the oracle software (both db and crs) at the same time.
The easiest part was to copy all the patches to the first compute node. Beside the CRS patch , you can start all the patches from a single compute node.
We planned a nonrolling patch.
We have started with Patch 9560930 – IB Switch software version 1.1.3-2 which upgrades the switches without any problem. Then we continue with Patch 10143395 – Cell patch 22.214.171.124.0 on the storage cells. This process is started from the compute node with patchmgr script.
Meantime we have started the Patch 10114705 – Exadata Bundle Patch 6 on the compute nodes, and after that we have passed Patch 10110978 – Exadata GI/CRS MLR4 on top of it.
On the storage site we had a problem on one of the cells. After the reboot, it alerts a file corruption on the root disk. We had run the checkdisk and the problem is fixed. But the cell auto rollbacked to the old version. At that time we had already finished the patches on the db side and started the db while 13 of the storage nodes on new version while the last one was on old version. We then started a rolling patch only for the remained cell. It ended without a problem.