"Strange, John W" <john(dot)w(dot)strange(at)jpmchase(dot)com> wrote:
> During heavy writes times we get the checkpoint too often error,
> what's the real knock down effect of checkpointing too often?
The main concern is that it may cause an increase in disk writes,
possibly to the point of causing blocking while waiting for the
disk.
Generally bigger checkpoint_segments settings improve performance,
especially as shared_buffers is increased. There are some
counter-examples, particularly during bulk loads, which haven't
really been explained:
http://archives.postgresql.org/pgsql-hackers/2010-04/msg00848.php
That makes this an area where careful testing of your real workload
with different settings can be important, at least if you're trying
to wring that last ounce of performance out of your server..
-Kevin