I recently upgraded from Postgresql 9.0.10 to 9.2.1. I am now running into problems with Postgresql running out of
memoryduring large data operations, more specifically loading the OpenStreetMap data into the database. The load under
9.0went fine and there were no memory issues. This is on the exact same machine, same postgresql.conf, same everything
exceptfor the upgrade to 9.2. Initially the OOM killer was kicking in and killing Postgresql. Once I set
vm.overcommit_memory=2,Postgresql just reports it is OOM rather than being killed. It seems the Postgresql process
keepsusing up more and more memory until it eventually fails, almost as if there is a leak.
Are there any new 9.2 memory usage parameters I may have overlooked? Here are some parameters I have set that worked
fineunder 9.0:
max_connections = 100
max_locks_per_transaction = 100
effective_cache_size=8GB
shared_buffers=4GB
work_mem=8MB
maintenance_work_mem=4GB
synchronous_commit=off
checkpoint_segments=100
checkpoint_timeout=10min
checkpoint_completion_target=0.9
The machine has 12 cores (24 w/ HT), 24 GB RAM, and is running CentOS 6.3 64-bit with all of the latest updates
applied.As I mentioned, over time, the Postgresql processes keep increasing memory usage until all physical memory is
usedup, and the process then fails.