"Tom Lane" <tgl@sss.pgh.pa.us> writes:
> The added WAL volume should be pretty minimal, because only tuples that have
> gone untouched for a long time incur extra work.
That seems like a weak point in the logic. It seems like it would make VACUUM
which is already an i/o hog even more so. Perhaps something clever can be done
with vacuum_cost_delay and commit_siblings.
Something like inserting the delay between WAL logging and syncing the log and
writing to the heap. So if another transaction commits in the meantime we can
skip the extra fsync and continue.
--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com