On Mon, Jun 27, 2016 at 8:28 PM, Michael Paquier
<michael.paquier@gmail.com> wrote:
> Could you post it? I'd be interested in seeing that.
Not right now, because it involves mutated customer data. I can
whittle it down later -- it's probably just a matter of having the
sweetspot number of TOAST chunks, or something like that.
What I can provide right away is some pg_xlogdump output (well, SQL
query results against a table loaded with pg_xlogdump output),
including some rough annotations of my own, highlighting the
transaction that raises the error (the one that has an abort record,
since I didn't force a PANIC). That is attached.
--
Peter Geoghegan