Mastodon's Next Major Release Enables Full-Text Search

Sean Tilley@lemmy.mlmod to Fediverse@lemmy.ml – 380 points –
429 Too Many Requests
wedistribute.org

It's hard not to overstate how much of a big deal it is that Mastodon is adopting this kind of search functionality. Mastodon still makes up a vast portion of the Fediverse.

While other platforms have supported this for way longer, having buy-in by the biggest player in the space will probably have a huge effect on standard expectations moving forward.

17

You are viewing a single comment

Let's see how many servers actually implement ElastiSearch. It's kinda resources-heavy.

Yeah, I feel like they'd probably get better mileage with something like Meilisearch, which is what Firefish uses for search.

It also adds to the deployment complexity even more. Just from memory, to run Mastodon you need:

  • any number of Rails web servers (horizontally scalable)
  • any number of Sidekiq worker processes (horizontally scalable)
  • a PostgreSQL database for persistent storage (vertically scalable modulo sharding)
  • a Redis server for caching and Sidekiq (vertically scalable modulo sharding)
  • a Elasticsearch server for full text search (vertically scalable modulo sharding)

So this is at least 5 different server processes to manage, In reality for almost all deployments, Redis and Elasticsearch are unnecessary; the database can be used for jobs and full text search. Further, it could even be SQLite for all but large instances.

The deployment story for Mastodon is a nightmare and a substitute like Pleroma or even better something in Rust is necessary.