On Wed, Mar 7, 2018 at 3:50 PM, Alvaro Herrera <alvherre@xxxxxxxxxxxxxx> wrote:
>
> Scott Frazer wrote:
>
> > It's only happening on the read replicas, though. I've just set my master
> > to handle all the traffic, but that's not really sustainable
>
> I failed to notice at start of thread that these were replicas. I
> suppose then you would have to be thinking about the range of XIDs in
> the primary server. Maybe you left some old transaction open, or an
> uncommitted prepared transaction there?
>
> If not, I suppose you're up for some "perf" profiling ...
>
> Scott Frazer wrote:
>
> > It's only happening on the read replicas, though. I've just set my master
> > to handle all the traffic, but that's not really sustainable
>
> I failed to notice at start of thread that these were replicas. I
> suppose then you would have to be thinking about the range of XIDs in
> the primary server. Maybe you left some old transaction open, or an
> uncommitted prepared transaction there?
>
> If not, I suppose you're up for some "perf" profiling ...
I think you got it. I found a 2-day old transaction on the master server that never closed. I had to pg_terminate_backend on it and the problem seems to have cleared up.
