Johnny Tan <johnnydtan@gmail.com> wrote:
> Wouldn't this be controlled by our checkpoint settings, though?
Spread checkpoints made the issue less severe, but on servers with
a lot of RAM I've had to make the above changes (or even go lower
with shared_buffers) to prevent a burst of writes from overwhelming
the RAID controllers battery-backed cache. There may be other
things which could cause these symptoms, so I'm not certain that
this will help; but I have seen this as the cause and seen the
suggested changes help.
-Kevin