Re: Frosen logical replication - Mailing list pgsql-general

From Ron Johnson
Subject Re: Frosen logical replication
Date
Msg-id CANzqJaANYmsHk77D4v3Hefa=77PUBD3quxMeHb5rR+cApHbrcA@mail.gmail.com
Whole thread Raw
In response to Re: Frosen logical replication  (Marcos Pegoraro <marcos@f10.com.br>)
List pgsql-general
On Fri, Dec 5, 2025 at 11:48 AM Marcos Pegoraro <marcos@f10.com.br> wrote:
Em sex., 5 de dez. de 2025 às 09:03, Marcos Pegoraro <marcos@f10.com.br> escreveu:
I have a logical replication where I want to replicate only one schema.
All worked fine, it copied all tables to subscriber, except one. That table has 8GB and it copied 5GB. That copy has been frozen since yesterday, as you can see.

I tried these steps.
- Tried to restart subscriber, didn't solve and file didn't change its size on subscriber. 
Subscriber did not try to restart copying from scratch, just did nothing.
DIDN'T SOLVE

- Tried to restart publisher 2 or 3 times, didn't solve the problem but file changed 
some MBs but was copying from what starting point if server was restarted and 
file was partially transfered ? I don't know. Then on subscriber side I got some messages like 
"could not receive data from WAL stream: SSL error: unexpected eof while reading"
DIDN'T SOLVE

- Restarted publisher and while restarting I truncated that table on subscriber. 
How srsubstate was "d" but srsublsn was NULL It had to copy that file entirely, and it did.
Then that file was completely copied from publisher and until now everything seems fine.
SOLVED, APPARENTLY

Maybe have multiple replication slots, one per X number of customers?  More work for you, but less work required by that single-threaded publisher, and more granular: hopefully most slots will replicate.

--
Death to <Redacted>, and butter sauce.
Don't boil me, I'm still alive.
<Redacted> lobster!

pgsql-general by date:

Previous
From: Marcos Pegoraro
Date:
Subject: Re: Frosen logical replication
Next
From: Thomas Munro
Date:
Subject: Re: wdavdaemon / Microsoft Defender for Endpoint on Linux and slow Postgres recovery?