Sergey Prokhorov

> I suspected it had to do with the SELECT queries I'm making in create_replication_slot checking to see if a slot already exists. Oh, good point! Looking in https://github.com/epgsql/epgsql/blob/devel/doc/streaming.md, it...

Well, at the very beginning of this page https://www.postgresql.org/docs/current/protocol-replication.html it says: > ... backend to go into physical replication walsender mode, wherein a small set of replication commands, shown below,...

Well, as you might notice, the process that printed this scary error report is neither 361 nor 365... What is this "" process? Why there is a bunch of other...

I honestly don't think I'll have time to dig more into this. I'm almost sure it's somehow related to the fact that you do this `select` before `start_replication`, but in...

Hi, sorry for late reply! Missed your report while being on vacation. Is this problem only reproducible while using `wal2json` or with other output plugins as well? Is it only...

Hi. There is no such plans. But it looks like it can be easily implemented as smth like ``` connect([ServerOpts | Servers]) -> try epgsql:connect(ServerOpts) of {ok, _} = Res...

So, I think you can use the schema that I suggested in 1st comment: ``` {ok, C} = connect([#{host => "master.host.name", ...}, #{host => "standby.host.name", ...}]) ```

> in example in the case of a network split, you can be connecting to a standby server like if it was a master. But does jdbc failover works differently?

Generic SCRAM implementation, but it uses NIFs https://github.com/esl/fast_scram/

What is the size and other parameters of your connection pool? I may imagine this error to happen when you open and close connections to your database very often (hundreds...