The way CP flu****ng is done has changed repeatedly over the years. Way
back
in the bad old days, a consistency point flush used to stall the server.
Also, the WB threads used to help out with CP flushes. Later the
behaviour
was changed so that CP flushes no longer stall the server and the WB
threads
no longer assist. I can't pretend to follow its vagaries from rev to rev.
I am suddenly curious to know a bit more about CP flu****ng in Ingres r3.
Specifically, how high a priority does CP flu****ng have? When I see two
CP
flushes of similar numbers of buffers using a similar number of DIOs,
should
I expect to see the duration of the flush vary by a factor of up to 10?
Roy
