At small scale, your database's built-in full-text search is often enough. Past that you add a dedicated search engine, and the key rule is: the index is a derived copy. The primary database stays the source of truth, and the index can always be rebuilt from it.

To keep the copy fresh:

save the post in the primary database, and record a "post changed" event in the same transaction (an outbox)
an indexer worker reads the events and updates the index. Updates and deletes must both flow through
the index is eventually consistent: a new post may take a few seconds to become searchable
to change the analysis or the schema, build a new index from scratch beside the old one, then switch over

A big index is split into shards. A query goes to every shard, each returns its best results, and the results are merged: scatter, then gather.

Find a word in a million posts