Re: Bug in ALTER SUBSCRIPTION ... SERVER / ... CONNECTION with broken old server - Mailing list pgsql-hackers

From Chao Li
Subject Re: Bug in ALTER SUBSCRIPTION ... SERVER / ... CONNECTION with broken old server
Date
Msg-id DFBCC95A-8B02-47E0-8825-9E00B51FB786@gmail.com
Whole thread
In response to RE: Bug in ALTER SUBSCRIPTION ... SERVER / ... CONNECTION with broken old server  ("Hayato Kuroda (Fujitsu)" <kuroda.hayato@fujitsu.com>)
List pgsql-hackers

> On Apr 22, 2026, at 20:35, Hayato Kuroda (Fujitsu) <kuroda.hayato@fujitsu.com> wrote:
>
> Dear Chao,
>
>> I tested the v19 new feature CREATE SUBSCRIPTION ... SERVER yesterday, and
>> found an issue: once the old server becomes broken, the subscription cannot be
>> recovered by switching it to a good server.
>
> Thanks for testing. I could reproduce the same issue. In addition to yours, I found
> DROP SUBSCRIPTION cannot be done anymore. To switch the connection or drop it,
> I had to create the same user mapping must be created again.
>
> ```
> postgres=# DROP SUBSCRIPTION sub_bug ;
> ERROR:  user mapping not found for user "postgres", server "old_srv"
> postgres=# CREATE USER MAPPING FOR CURRENT_USER SERVER old_srv
> OPTIONS (user 'dummy', password 'dummy');
> CREATE USER MAPPING
> postgres=# DROP SUBSCRIPTION sub_bug ;
> DROP SUBSCRIPTION
> ```
>
> Before deep dive to your fix, I'm unclear why dropping the active USER MAPPING is
> allowed. Personally, it should be avoided anyway. Do you know why it's not restricted?
>
> Best regards,
> Hayato Kuroda
> FUJITSU LIMITED
>

Hi Hayato-san,

There is an existing test case in subscription.sql:
```
-- fail, must connect but lacks USAGE on server, as well as user mapping
DROP SUBSCRIPTION regress_testsub6;
```

So, I guess that’s an intentional behavior. You have to fix the broken server or switch to a good one before dropping
thesubscription. That’s my understanding from the test cases. 

Best regards,
--
Chao Li (Evan)
HighGo Software Co., Ltd.
https://www.highgo.com/







pgsql-hackers by date:

Previous
From: John Naylor
Date:
Subject: Re: GUC parameter ACLs and physical walsender
Next
From: David Rowley
Date:
Subject: Re: New vacuum config to avoid anti wraparound vacuums