A dang week to build that activity visibility index. Again, pretty dang beefy box (24 cores/2x12, 32GB RAM, hardware RAID, etc.) with almost no load on it, real hardware rather than a VM.
On my desktop machine, the same dump took 12 hours to load. Like, literally 1/15th the time. Main difference is Postgres 12 instead of 9, which is a little baffling because 12 *should* be faster. Here's a screenshot of the time required, with units output overlaid, plus the previous screenshot of my extreme exasperation.
I think what I'm gonna do is just skip that index when loading (by means of grep -v) for local dev stuff and use the internal PG format for dumps so that everything can get loaded piecemeal, and for the big migration, see I can trick it into working in parallel. I haven't read the migrations for current Pleroma, I figure that index might have been changed by now (or I hope so), but in any case, this one's gonna require some thought.
180:32:29.pngthis_index_pushing_6_days.png