>>> I am running a site using Windows 2008 32-bit server and Firebird Database >>> Version 2.5.1. >>> >>> We have a mix of client connections utilizing direct access to the database >>> service as well as simultaneous Oracle ODBC sessions. There are 9 separate >>> database files for each of the connecting agencies. >>> >>> We are occasionally seeing the FBSERVER.EXE process spike to 50% constant >>> CPU usage, and no clients (or us using EMS SQL Manager Lite 2010 for >>> Interbase/Firebird) can connect to any of the databases. >>> >>> The client connections just hang once the Server process is stopped, they >>> all timeout/error out, and we start the service back up, and everyone can >>> get back in just fine. >>> > > At 07:43 a.m. 19/02/2013, Thomas Steinmaurer wrote: > >> A wild guess is that fbserver.exe (SuperServer) is busy doing something >> (costly report/processing, long-running query, sweep etc.) and by nature >> can only utilize max. one CPU/core per database. During that time, >> others are stuck or even can't connect. I have seen this many times in >> various support cases now with SuperServer. >> >> Switching to Classic (SuperClassic) and some fine-tuning usually helped >> a lot, especially on a SMP server. > > There's another resource aspect in this case, too: SS is a single process > running on a 32-bit server so it's limited to 2GB total memory. So, with "9 > separate database files for each of the connecting agencies" and an unknown > number of connections to each database by each agency, there will be a > tipping point where there's not enough RAM to make a new connection.
Could be, but as he is using SuperServer, the number of connections isn't in the equation, so it's the sum of page caches per database. -- With regards, Thomas Steinmaurer http://www.upscene.com/
