On Tue, 26 Feb 2008 03:25:46 -0500 [EMAIL PROTECTED] wrote: > Hi, > > I have been using mongrel + Pound for the past 6 months on a single server > deployment. > > Recently mongrel has started to behave very weirdly after we?moved app to a > bunch of servers from a single server deployment.
These answers might help: > - It crashes on every exception, throws 500 and throws 500 for ALL instances > of the cluster? after one exception has occurred For the crashes and database problems, make sure you have the mysql or postgresql gem installed and that it's using the actual binary library not the Rails built-in stuff. The built-in driver for those databases is bad and causes exceptions, lost connections, etc. > - Connectivity to SQL seems to take a rather long time (the db is a different > server on the same network, servers are connected through a gigaport switch) Rule out your network config before you assume it's a *Rails* issue (not a mongrel issue). Try running a few simple scripts and test the perf, then start from the most basic network config like half-duplex and see what you can find. > - Images deliveries have slowed down since we migrated from single server to > multi server deployment (still using pound, not yet moved to apache / nginx) Yeah, time to upgrade to nginx. Nginx is actually easier to configure than pound, makes more sense in the config, and is a hell of a lot faster and stable. Especially for having it serve static files. Just drop pound. > - The application architecture uses lot of background processes?- built > using?DRB which again are deployed physically on different dedicated machines > to distribute load Yeah, this really makes me wonder what you're doing. I'd go and double and triple check every line of code and watch network traffic to make sure you aren't doing something dumb like sending huge files over DRb. I'm guessing you've got something dumb that's overloading the Ruby interpreter. > - ?Mongrel doesn't always log the error 500 within its logs (mongrel may not > be getting these requests). Mongrel doesn't log most stuff, rails does. Check in the production.log file and see what's there. > -? Other mongrel instances in the same cluster do not respond even if one > mongrel has crashed or has encountered an exception. That sounds like Pound more than anything. Isolate the difference between a connection to pound and one to mongrel. > Here is some basic info of the deployment With this much hardware you should be able to crank tons of requests through the system. If you have this many problems, it's entirely in your design. Remember, tons and tons of other people do way more with much less. -- Zed A. Shaw - Hate: http://savingtheinternetwithhate.com/ - Good: http://www.zedshaw.com/ - Evil: http://yearofevil.com/ _______________________________________________ Mongrel-users mailing list Mongrel-users@rubyforge.org http://rubyforge.org/mailman/listinfo/mongrel-users
