Database Software

Hi All,

I was just wondering, what happens in a RAC setup when (let's say) hundreds of processes (on different processors) are getting the next value from a single sequence?

The reason that I ask is that I was trolling around on the web reading up on stuff and I came across an interesting comment where someone complained that using RAC, their application constantly gave unique constraint violations. I presume this was caused by attempts to insert duplicate PKs. The writer attributed this problem to a 7 second delay in propagating the cache bwteen nodes.

I have no experience with RAC myself, and am just in the research stage, but this seems to me like a problem with that application, not RAC. I can't imagine oracle building a system where sequences (the recommended, scalable way to get unique IDs) perform this way, so I have to presume that the said application isn't using sequences.

Any comments?
regards
Iain