Comment

Comment by bryantraywick

Others have replied with basically the same thing but replication has been in MySQL for a long time. Just last night I set up replication for a single MySQL server to a new 3-node Galera cluster then switched over the Galera cluster as the primary with only a few seconds of downtime. The downtime could have been entirely eliminated if the standalone MySQL server had binary logging enabled already.

Galera cluster is an awesome multi-master solution and I've had nothing but excellent experience with it in production environments. We currently typically direct writes at a single node and spread reads in round-robin to all three nodes using an HAProxy front-end to the cluster, but there is nothing preventing us from spreading writes across the cluster if it was needed. If the current write node goes down there is only a couple seconds of interruption as a new write node is elected and the application reconnects. Furthermore, if a node goes down rejoining is completely automatic once it starts back up with no slowdown in reads and writes. I don't have near as much PostgreSQL experience but replication and failover have been serious pain points from an operations stand-point when I have had to use it. I appreciate the end-user features of PostgreSQL, but operationally MySQL/MariaDB (plus Galera cluster) are miles ahead of anything I've seen for PostgreSQL.

parent bryantraywick 12 years ago

No comments yet

Be the first to share your thoughts.

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton