I originally drafted this article in 2015, when I was managing a production MariaDB Galera cluster for Stockopedia. It’s been gathering dust in my drafts since then. I don’t remember why I didn’t complete it; maybe the subject needed more treatment, maybe I couldn’t find a proper conclusion, or maybe I just lost interest. I don’t know. It’s Jan 2020, I’m finally pushing it out so I can call it done. This article reflects my learning from that period. I have since moved on, both from active duty at Stocko and managing databases in general. The state of the affairs may have improved since, I haven’t kept up. FYI
Intro
At Stocko I wear many hats, one involves running a MariaDB Galera cluster as our primary RDBMS. We came to Galera after running a MySQL async replica for a while. Our rationale was to start with what’s built-in, as that is the path of least resistance. In time, we knew we needed a better way to guarantee cluster reliability. After some research, it became clear Galera was the only active, open-source solution that could do multi-master replication.
Read more