Blue)(Fusion
09-02-2007, 12:20 PM
Well my employers had servers here for some time, but since they hired me on to do work for them, I have been very unimpressed with the operations at SF. I won't list out everything, but the latest issue has me quite tense.
So our main web server is getting overloaded. I used Lighttpd + mod_cache as a proxy on the server to interface with Apache on localhost which helped increase speeds somewhat, but there's still just a limit as to what it can do. To get thing going fast again we opted for a load balanced solution. The second web server and load balance server (they use Linux Virtual Server (LVS)) was brought online last week. The plan was to migrate everything to use the load balancer over this long weekend so we can have ample time to migrate and monitor the setup.
To accomplish this migration to the load balancer, we need SF staff to change the VLAN on the switch for our original web server to put it on the same IP block as the new units. I can set the server to the new IP, but without them changing it on the switch, it's useless. I was told that the guy at SF who does that is gone for the weekend.
In addition to not getting the server IP changed, we need the LVS software configured. We were told it was done, but when I logged in, I couldn't find the lvs.conf file (using slocate and find) so I am unsure if it's even properly installed. The ONLY guy who knows how to use the LVS software at SF is also gone for the weekend.
So now, our plans to make use of the free time we had this weekend ends up being a waste of our time. Maybe it's just me, but I find it ridiculous that the staff at SF rely on one or two individuals for certain tasks and if they're off duty, you're SOL. It would have been faster if I colocated my Foundry ServerIron load balancing switch there and have them attach our servers to it. If Meigs Field was still open, I could have flown to Chicago and set it up myself in a few hours.
That's all I have for now.
So our main web server is getting overloaded. I used Lighttpd + mod_cache as a proxy on the server to interface with Apache on localhost which helped increase speeds somewhat, but there's still just a limit as to what it can do. To get thing going fast again we opted for a load balanced solution. The second web server and load balance server (they use Linux Virtual Server (LVS)) was brought online last week. The plan was to migrate everything to use the load balancer over this long weekend so we can have ample time to migrate and monitor the setup.
To accomplish this migration to the load balancer, we need SF staff to change the VLAN on the switch for our original web server to put it on the same IP block as the new units. I can set the server to the new IP, but without them changing it on the switch, it's useless. I was told that the guy at SF who does that is gone for the weekend.
In addition to not getting the server IP changed, we need the LVS software configured. We were told it was done, but when I logged in, I couldn't find the lvs.conf file (using slocate and find) so I am unsure if it's even properly installed. The ONLY guy who knows how to use the LVS software at SF is also gone for the weekend.
So now, our plans to make use of the free time we had this weekend ends up being a waste of our time. Maybe it's just me, but I find it ridiculous that the staff at SF rely on one or two individuals for certain tasks and if they're off duty, you're SOL. It would have been faster if I colocated my Foundry ServerIron load balancing switch there and have them attach our servers to it. If Meigs Field was still open, I could have flown to Chicago and set it up myself in a few hours.
That's all I have for now.