One further question, how exactly can a CachedPlan leak? The OpenStreetMap
data load does use plsql and triggers are involved as well. Is there
something new to 9.2 that changes the way CachedPlans are cleaned up?
--
View this message in context:
http://postgresql.1045698.n5.nabble.com/Postgres
In the logs, after I got the OOM, I noticed I had several thousand lines
similar to the following:
CachedPlan: 7168 total in 3 blocks; 2472 free (0 chunks); 4696 used
CachedPlan: 7168 total in 3 blocks; 3200 free (0 chunks); 3968 used
CachedPlan: 1024 total in 1 blocks; 64 free (0 chun
