Euler Taveira

The only explanation for the OIDs in `Relation` to be invalid in following code is due to an invalidation. https://github.com/eulerto/wal2json/blob/75629c2e1e81a12350cc9d63782fc53252185d8d/wal2json.c#L2453-L2462 The `Relation` seems to be invalidated during this code path...

See [RelationCacheInvalidate](https://github.com/postgres/postgres/blob/master/src/backend/utils/cache/relcache.c#L2978-L3014). You should also check [inval.c](https://github.com/postgres/postgres/blob/master/src/backend/utils/cache/inval.c). Logical replication processes some invalidation messages when calls [AcceptInvalidationMessages](https://github.com/postgres/postgres/blob/master/src/backend/utils/cache/inval.c#L806). This function removes some catalog caches and relation cache that are not useful anymore....

The installations instructions were updated by efee43f09864b37192d4c63db0911b38321debd2.

What's the JDBC version? Is it a specific transaction that causes it? If so, could you share it?

There isn't enough information to say if it is a bug or not. Having said that, I'm closing this issue because we couldn't reproduce it and the reporter also didn't...

@hash-data This does not seem like a wal2json issue. The replication slot is not being consumed (_active = false_). Check the application that is responsible to start the logical replication.

The error messages need improvement. They are not informative enough to rapidly fix a parsing issue. You provide the line number but maybe it is a good idea to also...

You are almost there... As I said it should parse the whole file and then apply changes. Your patch is still applying changes while parsing the file (see `unquote_add_authfile_user` inside...

@devrimgunduz PR #55

Unless you can show that the extra time is spent on PgBouncer side, there is no bug here. I advise you to check hardware resources too.