Our inserts and updates on an older 7.3.4 cluster are very slow
(0.3s-0.9s) for any/all tables, new and old. I know an upgrade
may be in order, but I have a number of other 7.3.4 legacy
clusters, and I'd really like to understand the cause and if an
upgrade is going to solve this problem (downtime is sensitive,
of course).
As a test case, I created the following table on both the slow
cluster and a newly created test cluster on the same box:
create table foo(id integer, msg text)
I measured less than 10 inserts/second max on the slow cluster.
On the new cluster on the same system, I measured 5000
inserts/second. So I figure it's something in the system
relations. Autovac had been disabled for a few days at least.
I did a db-wide vacuum/analyze, still no speedup.
So two questions: 1) Any ideas on what is the cause? 2) How is
this avoided in 7.4.8 or 8.0.3?
Ed