Em seg., 21 de out. de 2019 às 21:20, Jeff Janes <jeff.janes@gmail.com> escreveu: > > After setting up logical replication of a slowly changing table using the built in pub/sub facility, I noticed way more network traffic than made sense. Looking into I see that every transaction in that database on the master gets sent to the replica. 99.999+% of them are empty transactions ('B' message and 'C' message with nothing in between) because the transactions don't touch any tables in the publication, only non-replicated tables. Is doing it this way necessary for some reason? Couldn't we hold the transmission of 'B' until something else comes along, and then if that next thing is 'C' drop both of them? > That is not optimal. Those empty transactions is a waste of bandwidth. We can suppress them if no changes will be sent. test_decoding implements "skip empty transaction" as you described above and I did something similar to it. Patch is attached.
Thanks. I didn't think it would be that simple, because I thought we would need some way to fake an acknowledgement for any dropped empty transactions, to keep the LSN advancing and allow WAL to get recycled on the master. But it turns out the opposite. While your patch drops the network traffic by a lot, there is still a lot of traffic. Now it is keep-alives, rather than 'B' and 'C'. I don't know why I am getting a few hundred keep alives every second when the timeouts are at their defaults, but it is better than several thousand 'B' and 'C'.
My setup here was just to create, publish, and subscribe to a inactive dummy table, while having pgbench running on the master (with unpublished tables). I have not created an intentionally slow network, but I am testing it over wifi, which is inherently kind of slow.
