Hacker News new | past | comments | ask | show | jobs | submit login

Tailing the WAL is an interesting approach. How do you handle the potential increased load on the database from constant WAL reads?



We use PG replication slots and listen to updates. This doesn't add much load to the database -- it's similar to having a read replica. Adding more servers would mean more replication slots, and could slow down PG. When this happens, we'll likely replicate PG's WAL onto Something like Kafa. This is what LiveGraph does [1]

[1] https://www.figma.com/blog/livegraph-real-time-data-fetching...




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: