We have a relatively small dataset loaded and have 3-4 people editing simultaneously. We aren't doing a bulk upload.
Everything is on one box, but it's a decent box: 4 fast CPUs, 6GB RAM, SATA RAID. I just did some tweaking. Set PasengerMaxPoolSize 20. Increased SHMMAX so I can increase Postgres' buffer size. I am now seeing memory usage approach 100% (but not go to swap). But it still seems to bog down after a few minutes of intense editing with Potlatch (edit with save). Vector data is slow to return when scrolling (getting the blinky triangle). -Eric -=--=---=----=----=---=--=-=--=---=----=---=--=-=- Eric B. Wolf 720-334-7734 On Mon, Sep 13, 2010 at 3:49 PM, Brendan Morley <[email protected]>wrote: > What parts of the API are you exercising? > > When setting up servers for CommonMap, I found that a 50,000 element upload > takes 40 to 80 minutes to complete, depending on the type of server I used. > > Ruby on Rails is very inefficient for that kind of thing. > > > Brendan > > > On 14/09/2010 6:15 AM, Eric Wolf wrote: > > We finally have our system running and are doing some field tests with > multiple editors. The system keeps getting bogged down. Looking at the > process list, I'm seeing 7-8 ruby tasks, each eating 50%-100% of a CPU core > (thank god for multiple CPUs & cores). They are persisting for 20+ minutes. > Any idea what may be up? Should ruby tasks be hanging around that long? > > I'm running a pretty much default Postgres install. Any hints on > optimizing Postgres for OSM? > > -Eric > > -=--=---=----=----=---=--=-=--=---=----=---=--=-=- > Eric B. Wolf 720-334-7734 > > > > > _______________________________________________ > dev mailing > [email protected]http://lists.openstreetmap.org/listinfo/dev > > > > _______________________________________________ > dev mailing list > [email protected] > http://lists.openstreetmap.org/listinfo/dev > >
_______________________________________________ dev mailing list [email protected] http://lists.openstreetmap.org/listinfo/dev
