Re: [PERFORM] Re: bgwriter autotuning might be unnecessarily penalizing bursty workloads

2013-07-18 Thread Jeff Janes
On Thu, Jul 18, 2013 at 10:40 AM, Josh Berkus wrote: > > Right, that's what I just tested. The results are interesting. I > changed the defaults as follows: > > bgwriter_delay = 100ms > bgwriter_lru_maxpages = 512 > bgwriter_lru_multiplier = 3.0 > > ... and the number of buffers being written by

Re: [PERFORM] Re: bgwriter autotuning might be unnecessarily penalizing bursty workloads

2013-07-18 Thread Josh Berkus
Greg, > There look to be a good number of buffers on this server that are only > being written at checkpoint time. The background writer will only deal > with buffers when their usage count is low. Fast servers can cycle over > shared_buffers such that as soon as their usage counts get low, they

[PERFORM] Re: bgwriter autotuning might be unnecessarily penalizing bursty workloads

2013-07-17 Thread Greg Smith
On 7/17/13 6:46 PM, Josh Berkus wrote: The key metric I'm looking at is that the bgwriter only took care of 0.3% of buffers. There look to be a good number of buffers on this server that are only being written at checkpoint time. The background writer will only deal with buffers when their u