Re: resolving buffer busy waits

late update, but thankfully a positive one.

series of quick outages over the weekend has removed the problem. main issue was setting disk_asynch_io to false and adding 6 db_writer_processes. other changes involved the usage of a large keep buffer pool and some adjustments to checkpoint parameters.

had to wait until this evening when weekly activity peaks to really "test" the system under load. all the red herring waits were gone -- db file sequential read, buffer busy waits and enqueues. all just hiding the underlying problems.

if anyone wants more info, email me direct.

:-) Received on Mon Sep 22 2003 - 15:37:22 CDT