-
Hi, |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 26 replies
-
Heya! For 1&2: Because we use postgres' logical replication, combined with replication slot, pg_flo only sees committed transactions. The operations are then sent in order to NATS. From there the worker also pulls them in order, applies any transformation or rules and then syncs to the appropriate, all in the same order. If a transaction wasn't committed in postgres or rolled back, pg_flo won't see that. Also, for related reasons, it's recommended to only have one worker for now - https://github.com/pgflo/pg_flo?tab=readme-ov-file#scaling-guide. It can be 5-8k ops/s. For 3: Thats right, one group is basically equivalent of one Publication and replication. Its basically a grouping of all the tables that are being replicated. Let me know if you have any other questions or feedback, thanks! |
Beta Was this translation helpful? Give feedback.
Heya!
For 1&2: Because we use postgres' logical replication, combined with replication slot, pg_flo only sees committed transactions. The operations are then sent in order to NATS. From there the worker also pulls them in order, applies any transformation or rules and then syncs to the appropriate, all in the same order. If a transaction wasn't committed in postgres or rolled back, pg_flo won't see that. Also, for related reasons, it's recommended to only have one worker for now - https://github.com/pgflo/pg_flo?tab=readme-ov-file#scaling-guide. It can be 5-8k ops/s.
For 3: Thats right, one group is basically equivalent of one Publication and replication. Its basically a grouping of all t…