In a Web cluster, be sure to ensure that each machine program code file is the same.
So the general two scenarios,
One is to synchronize the uploaded code to each server by synchronizing the software, that is, each machine really has its own code.
The second is for each server to mount a common file storage server.
Both of these scenarios, the second one is obviously easy to manage.
The first possibility is better when the site is running, because the native files are read directly. But in the management inconvenience, such as logging in the program, so many servers have their own logs, to manage these logs, but also to merge, and not a simple merger, multiple servers the same name log file on the same day every log time has a cross, ordered to do a log analysis management.
One way is to store the logs separately for the co-mounted server, and the Web site's file upload business does the same thing.
But the site project is more, programmers use different language framework, log directory configuration is inconsistent, so that each program to manually save the directory.
I prefer the second, that is, the way each web is mounted together, but is there a problem with the server having dozens of hundred large clusters?
Are the code files running on hundreds of Web servers co-mounted to a file server?