We have a group of four web apps (C#, ASP.NET, NHibernate) and a few services that used to share a Northscale instance. We've since upgraded to Membase, at which point we were actually able to start making effective use of the NHibernate second level cache. However, once we got the apps into production, the memcache service stopped listening to telnet on port 11211 and the management interface showed that it was using 9.99k connections. Judging by the traffic slamming into our database, we believe that one or more of the following steps should occur. In both cases, the server was located on a Windows 2008 server.
1) We should allow more connections to the server. This is allegedly done by calling the executable with the -c parameter. In practice, on our machine, we get an error indicating that it can't load default_engine.so when we attempt this. 2) We should decrease the pool size that our apps are using. Currently, the main web app (which has many instances that run for different clients), has a pool sized between 10 and 100 connections. 3) We need to consider whether it was an issue of network saturation. According to the management console, we had more than enough RAM. Is it possible that all the traffic overwhelmed some of the hardware at Rackspace?
Is there any other obvious source of problems that we could have overlooked?