Hi,
We are migrating to a new server with more memory and also from
postgres 7.4 to postgres 8.0.1 version.
Here are my settings on the current 7.4 version:
OS : RedHat 9
CPUS: 2 hyperthreaded
Memory: 4gig
shared_buffers: 65536
sort_mem: 16384
vacuum_mem: 32768
wal_buffers: 64
effective_cache_size: 393216
checkpoint_segments: 3
checkpoint_timeout: 300
checkpoint_warning: 30
These are settings which I am planning on the new machine with 8.0.1
version:
OS: Fedora Core 2
CPUs: 2 hyperthreaded
Memory: 8 gig
shared_buffers: 131072
work_mem: 32768
maintanence_work_mem: 65536
wal_buffers: 64
effective_cache_size: 786432
checkpoint_segments: 8
checkpoint_timeout: 600
checkpoint_warning: 30
The current settings on my 7.4 version gives me very good
performance, so I basically doubled the settings since i will be having
the double the memory in the new machine. What my concern is about the
effective_cache_settings , according the docs its recommends me to set
max to about 2/3 of the total memory and I went little over on top of
it, is that ok ? I went little over on my current 7.4 system too, and
its giving me very good performance so I used the same calculation for
my new system too.
Also, can anyone guide me with the ideal settings for
vacuum_cost_delay, vacuum_cost_page_hit, vacuum_cost_page_miss,
vacuum_cost_page_dirty, vacuum_cost_limit, background_delay,
bgwriter_percent, bgwriter_maxpages settings. I am not sure what
settings should I make to these parameters , are there any ideal
settings for these parameters in a OLTP environment ?
Thanks!
Pallav