Re: BUG #14639: Different xmin values in a transaction

Hi Andres,

I care about xmin values because we have a very high number of inserts/updates/deletes and we get a Vacuum Freeze (from autovacuum) occurring every few days so I want to reduce that happening. The application is doing inserts in batches so I expected to see batches with the same xmin. When I found they were all different I then identified what I thought was an issue with the partition function. We can work around it but I wasn't expecting this behaviour.

Regards ... Ray

-----Original Message-----
From: Andres Freund [mailto:andres(at)anarazel(dot)de]
Sent: 02 May 2017 17:13
To: Ray Warren
Cc: pgsql-bugs(at)postgresql(dot)org
Subject: Re: [BUGS] BUG #14639: Different xmin values in a transaction

On 2017-05-02 15:57:39 +0000, ray(dot)warren(at)artsalliancemedia(dot)com wrote:
> Under certain conditions inserts in the same transaction are getting
> different xmin values.
> A reproducible example is below. I can work around it but I think this
> is a bug.

That's to be expected if you use savepoints (and EXCEPTION blocks use those internally). Why do you care about xmin value?

Greetings,

Andres Freund