Conversation
Notices
-
@lanodan @hj @get @j We had that disgustingly overstuffed 384GB, constant performance-tuning, Postgres had ~100GB dedicated (although it was difficult to get it to make use of that much and it usually hovered around 60GB). If the speed issue is a lack of RAM, it wouldn't have happened on FSE.
If I pop up the regular FE (keep in mind, we're on an old version, it may have changed), hit "@as" and then mash "d" then backspace repeatedly, it sends a request per keypress (even though it's the same two GETs), same 43kB response each time; might severely reduce server load across all Pleromae if there was just a delay, wait until they've stopped typing for 250ms before fetching the results and cache them (because someone using autocomplete as an ad hoc search feature is going to hit backspace a lot if the visible results don't include the expected name). And that's a req worker, a slice of the pipe, a connection checked out of the DB for about a second (in the current case, it's an ARM system with much less RAM dedicated to Postgres and fewer users, not really comparable, but 1400ms until the rate-limiter starts kicking the requests back).
- † top dog :pedomustdie: likes this.
-
@hj @j @get Also vast majority of admins not having enough RAM to store the search index, and not knowing they need it because they've never dealt with actual database load before.
-
@j @get @lanodan text search on massive databases difficult and slow
-
@get that's actually a good question
cc @lanodan @hj
-
why is pleroma search so bad
-
>accidentally hit unfollow on j
>instantly disappears from my feed competely
>pleroma search is useless, returns nothing when I search for @j
>try to hit remote follow on bae st, gives error
the fediverse user experience