Four virtual machines were running on my desktop. A fifth was still installing in the background.
The console reached them through a server at the colocation facility. I took a screenshot and sent it to the client because the whole setup was finally visible in one place.
The goal was to rebuild the Linux Virtual Server layer before touching production. LVS would sit in front of the web servers, distribute traffic, and fail over if one side stopped working.
I got the virtual version running with the same routing and failover pattern we planned to use for the live systems. Once that worked, I wanted to build the replacement web servers, put the application code on them, and have the client's team test everything within a day or two.
This was much safer than inventing the network directly on the production boxes. Virtual machines let me create the parts, break the routing, watch the failover, and start again without taking the customer-facing system down each time.
It was also a lot of infrastructure for one local desktop. Five machines, multiple networks, a remote colocation host, load balancing, heartbeat, and a production plan were all sitting behind one screenshot.
The subject of my next email was "LVS failover money." That was not a formal project name. It was how I wrote when the thing finally worked and I wanted to move.
Full routing. Failover. Same shape as production.
Now we could build the web servers and let the application team try to break them.