RE: ASM is single point of failure ?

>In a RAC env, you will have an ASM instance on each node, so there is
no SPOF (beside
>your disks, but they are usually mirrored, so the risk is lessened).

If you lose an ASM instance, all the database instances on that node are worthless. You must have a local ASM instance. This post makes it sound as though somehow any surviving ASM instance is sufficient and that is not the case.

This debate over ASM will get old soon. If you have no other way to get RAID 1 and 0, then use ASM.

On the other hand, Oracle has certifaction programs for such functionality as 3rd Party CFS (which generally have volume managers underlying). To see which ones are available on Linux, go here:

http://www.oracle.com/technology/products/database/clustering/certify/te ch_linux_x86.html

To see what the program is for this testing go here:

http://www.oracle.com/technology/software/oce/oce_fact_sheet.htm

If you ignore these certification programs and the technology listed there, then ASM is for you. If you want to use Oracle datafiles the way it has been done for 30 years (thus not "trying" new stuff), than go with a certified CFS.

In the end, however, stating on this list that you have a couple of RAC servers (with or without ASM) that are nice and stable is only a case in point and doesn't mean the mix is stable. It just means it has worked in your environment.

Not everyone on this list leaps to experiment with their production data, particular on any technology that fundamentally changes how Oracle has done storage for 30 years.

--
http://www.freelists.org/webpage/oracle-l