hhmx.de

· Föderation EN Mi 16.04.2025 17:23:44

While we have already restored the server from a hardware failure, we're having struggle reconnecting the node to our Galera (MariaDB database) cluster. The issue seems to be independent.

Föderation PL Mi 16.04.2025 17:31:08

@Codeberg what failed btw?

Föderation EN Mi 16.04.2025 17:36:54

We seem to have had luck manually recovering some internal files and initializing a sync. It will run for a while, but it looks promising.

Föderation EN Mi 16.04.2025 18:18:56

@Codeberg I'm just curious, what is the reasoning for having picked Galera over standard replication? Thanks

Föderation EN Mi 16.04.2025 18:26:23

@Techwizz What do you mean by "standard replication"?

Föderation EN Mi 16.04.2025 20:25:41

@Techwizz Hmm, no specific reason. Galera sounded like better suiting our needs (a three-node multi-primary cluster, because Forgejo has no builtin tooling to separate reads and writes so far).

Föderation EN Mi 16.04.2025 21:17:28

@Codeberg @Techwizz Nothing wrong with Galera as such. I used it in a heavy load production cluster for years without issues. The whole multi primary thing saved me more than once. It does come at a performance cost though. We also ran a large-ish production Openstack environment using Galera but there we used a single primary with haproxy in front for a combination of fail over and performance.

Föderation EN Do 17.04.2025 00:33:19

@Codeberg Cool, thanks! Was asking as I've been debating between Galera and standard replication for a prod environment at my work. Thanks for the answer!