On Tue, Jun 7, 2016 at 6:33 PM, Peter Tripp <peter@chartio.com> wrote:
> From my logs:
> ````
> 2016-06-08 00:54:18 UTC:10.10.4.142(59691):user@dbname:[13169]:ERROR:
> attempted to delete invisible tuple
> 2016-06-08 00:54:18 UTC:10.10.4.142(59691):user@dbname:[13169]:STATEMENT:
> INSERT INTO cache (key, value, expires_at) VALUES ($1, $2, $3) ON CONFLICT
> (key) DO UPDATE SET value=EXCLUDED.value, expires_at=EXCLUDED.expires_at
> ````
>
> This was a freshly provisioned postgreSQL 9.5.2 on RDS with no imported
> data. The issue appeared less than a minute after stress testing it with
> insert/updates. I don't yet have a working test case, but I've only seen it
> when there's quite a bit of concurrency.
This is the stress-testing suite that we used during development:
https://github.com/petergeoghegan/jjanes_upsert
Does your test case also use foreign keys? Or, is that just the other
case reported by Viren Negi?
Thanks
--
Peter Geoghegan