Hi my k3s peeps! I want to enable HA on a two node cluster. Is it possible to run MySQL inside the the cluster using a MySQL operator, and then still add other master nodes at that point? Or does MySQL need to be run outside of the cluster to solve some chicken-or-the-egg-first type problems?

you want to have the cluster run against a sql database, that is hosted inside the cluster? How would you get the database running before the cluster is up, if it is hosted inside the cluster?

This is why I'm asking the question 🙂 Not sure if there was some mechanism where you could start with the embedded sqlite, and then specify a new database (hosted inside itself) and it would work. If you guys have never heard of it then my suspicions were correct. I googled and LLM'd pretty extensively prior to this, just wanted to make sure we're not missing some easier management of a database by using an operator!

nooope

We are running two edge nodes, which is why we can't use etcd

Even with active/active SQL you still really want a 3rd node to act as a witness for proper HA

We have hundreds of clusters, and so far, they're all the same. It would be a huge and immense effort to deploy a third. And cost for those customers...

There isn’t really a good solution for proper two-node replicated HA

Yeah... We are coming from a KeepaliveD / HAProxy env. So obviously not 1:1, but k8s offers such a better environment for deploying services!

Is NATS still an option? Assuming the goal is 2-node HA. The next gen platform for the edge: SUSE and Synadia Bring Two-Node High Availability to Kubernetes | SUSE Communities

I was literally just about to ask if anybody has seen people use Marmot, which uses NATS underneath I believe.

no, nats got abandoned by the synadia team

Sadness.

That is, it seems like it could have been a pretty simple drop in solution.