Some time ago I wrote a blog about stretched clusters and the OCR. The final conclusion at that time was that there was no easy way to get your OCR safe on both storages, and hence I disrecommended clusters with 2 storage boxes. However, after some more investigation I may have to change my mind. I did extended testing on the OCR and in this blog I want to share my experiences.
This is the setup:
- 2-node RAC cluster (10.2.0.4 on Solaris), located in 2 server rooms
- 2 storage boxes, one in each server room
- ASM mirroring of all data (diskgroups with normal redudancy)
- One voting disk on one storage box, 2nd voting disk on the other box, 3rd voting disk on nfs on a server in a 3rd location (outside the 2 server rooms)
For the components above, this setup is safe against server room failure:
- The data is mirrored in ASM and will remain available on the other box.
- The cluster can continue because it still sees 2 voting disks (one in the surviving server room and one on nfs).
But what about the OCR?
We did as what looks logical: OCR on storage box 1 and OCRmirror on storage box 2, resulting in:
Device/File Name : /dev/oracle/ocr Device/File integrity check succeeded Device/File Name : /dev/oracle/ocrmirror Device/File integrity check succeeded
Now we can start playing. For the unattended reader, “playing” means: closing ports on the fibre switches in such a way that a storage box becomes totally unavailable to the servers. This simulates a storage box failure.
The result is a story of 5 chapters and a conclusion. Please standby for the next upcoming blog posts.