Re: [GENERAL] Odd performance issue

2005-06-20 Thread Tom Lane
Bart Grantham <[EMAIL PROTECTED]> writes: > ... it may be relevant that one node_id has 15 times as many connections: > ... but that shouldn't make a difference to the planner, should it? Of course it should (and does). > Yes, I've vacuum analyzed. What you need to do is show EXPLAIN ANALYZE o

[GENERAL] Odd performance issue

2005-06-20 Thread Bart Grantham
Summary: depending on the value, the planner will sometimes choose a seq scan, sometimes an index scan. The former produces terrible performace, the latter great performance. The long story: we had a disk failure (NOT the disk the db was on) and the machine's system disk had to be rebuilt fro