the writeup
Super Bartie, live
This is a Postgres CDC pipeline running on one VM. Changes go from the source's WAL through Redpanda into a second Postgres, and each row is also stored as a vector. The panels below call the running system.
terra → warehouse…
live · superbartie.duckdns.orglatency
…
reader lag
…
backlog
…
merge p95
…
Stops the writer so you can watch the backlog build. It resumes on its own.
1. Watch rows land
Newest rows on each side. A script adds a sighting every two seconds. Highlighted rows arrived in the last few seconds. Click a sighting to edit it in step 2. browse both databases
source (terra)
loading…
destination (warehouse)
loading…
2. Change a row on the source
observations #9001 · demo rowMove the sighting somewhere else. Pick a place and the change goes to the source.
reset goes back to demo row #9001 at south bank
source (terra)
∅
loading…
destination (warehouse)
∅
loading…
3. Ask the replicated data
pgvector · same question twicelive copy
updated by the pipeline, about two seconds behind the source
batch copy
a snapshot taken every five minutes