Thread: BUG #8984: ERROR: t_xmin is uncommitted in tuple to be updated

The following bug has been logged on the website:

Bug reference:      8984
Logged by:          Marko Tiikkaja
Email address:      marko@joh.to
PostgreSQL version: 9.1.11
Operating system:   64-bit Ubuntu
Description:

Hi,

I just noticed we've had the error in subject happen around ten times,
starting from Thursday.  I don't see it having happened before that.  IIRC
that day we also switched to another synchronous slave.  The previous slave
had hot_standby_feedback=on, and the new one has it turned off.  Might just
be a coincidence.

All of the failing queries are inside PL/PgSQL functions, most of them are
SELECT .. FOR UPDATE or UPDATE against a table called "workers", but there
are a couple of instances against other tables as well.

Any ideas?  I'll be happy to provide more information, just let me know what
you need.

Hi,

It looks like this could happen if the xmin of the tuple happens to be
the reported xid of a standby with hot_standby_feedback, no?  But I
would expect this problem to happen more frequently if that's true..


Regards,
Marko Tiikkaja

On 1/29/14, 12:25 PM, I wrote:
> It looks like this could happen if the xmin of the tuple happens to be
> the reported xid of a standby with hot_standby_feedback, no?

My bad, hot_standby_feedback sets xmin, but TransactionIdIsInProgress()
only looks at xid.  So that theory is obviously bogus.


Regards,
Marko Tiikkaja