There's something to be said for simplicity. So far, the tools we've discussed in this chapter are larger client-server mechanisms or components of entire toolkits. One of the central tenets of the Unix philosophy is to build tools that do one thing well. In this case, we turn to Peak6 and their walctl WAL-management tools.

I created walctl specifically to address shortcomings in existing WAL-related utilities. Primarily of note is the question of architecture. Existing WAL tools follow an architecture diametrically opposed to the end goal of high availability. We often see this:

In this kind of model, the master node is tasked with transmitting transaction streams or WAL files to every node in the cluster. This makes it fantastically difficult to change the active master node and potentially overloads the master node itself. The primary write node of any cluster should be focused on fulfilling client requests. The purpose of walctl is to impose a structure...