Hi,
On 2022-02-14 15:45:40 -0500, Tom Lane wrote:
> I tried to replicate the problem described in [1] about logical
> replication from a v14 source DB to a v11 target. It fails as
> described there; I've not yet tracked down why, but it looks like
> the v11 apply worker fails and closes the connection after sending
> CREATE_REPLICATION_SLOT. The v14 walsender doesn't have a problem
> with that. However, if I do the same thing using HEAD as the source,
> I see an assertion failure on the way out of the walsender. The
> postmaster log shows
Gah, sorry, forgot about this one. I'll fix it / apply the fix.
> I guess this indicates that walsenders don't initialize their pgstats
> support soon enough?
It's more that we don't release slots early enough, I think.
Greetings,
Andres Freund