> anyone using postgres 9.3+ with big buffer sizes? (Not sure where > big starts) >
We use Postgres 9.3 on SmartOS with 12G shared buffers. We configure all our Postgres instances to use the minimum of 0.25 available RAM or 12G for shared memory buffers. > On solarish system the shared mem is allocated as ISM. > (See bug report above) > I kind of wonder what performance impact (except for beeing locked > memory with ism) this has and if anyone every tested it. > > In general things work fine. Different memory reporting tools report different values, depending on how they track memory usage. If it looks as if all your memory is being consumed, try a few tools to verify before freaking out. If you have many connections, some memory accounting algorithms will think that you're consuming all available RAM. You may see an initial lock up of the system (for a few seconds) as it reads through all the memory pages to determine real usage, but the OS will cache the ratio of reported memory to real memory usage and behave normally after that. We've gone through a number of different postgres configurations to reach a happy stable state. Here's a copy of one, with a bit of identifying info removed: https://gist.github.com/sax/beff6ea44a0428a920c9. Definitely configure the ZFS block size to 8k, to match the default postgres block size. -- ---- e s ------------------------------------------- smartos-discuss Archives: https://www.listbox.com/member/archive/184463/=now RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00 Modify Your Subscription: https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb Powered by Listbox: http://www.listbox.com
