What about fail over?
Our new AsyncSocketAppender has built-in fail over mechanism. You can specify several log servers for your application to work with. If current connection gets lost, the appender will automatically switch to the next host in the list. The switching is performed on the background and should be transparent for your application. Usually re-connection takes some time, but we also do the queuing of the log events so that it's not lost during fail over. Once connection is re-established, queue is transmitted to the new server. Again, this is done asynchronously and shouldn't affect the application, accept, of course, the resources taken - the CPU and in-memory queue size, which is configurable. More details on using AsyncSocketAppender can be found in user manual.
On the other hand, you can set-up, for example, two identical logFaces servers so that they complement each other in case one fails. Instances of the server can safely share one database. When one server goes down, the appender will automatically switch to the second server and all information will be saved at the same database. Make sure to configure both servers with identical parameters so that you won't feel a difference when fail over takes place. Particularly important is to setup database capacity and filters to be the same in both instances.