Re: emergency outage requiring database restart

On Wed, Oct 26, 2016 at 1:09 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Merlin Moncure <mmoncure(at)gmail(dot)com> writes:
>> *) I've now strongly correlated this routine with the damage.
>
> Hmm. Do you have any way to replace the non-core calls with something
> else? The "shexec('rm -f ' || _OutputFile)" bits could presumably be
> converted to use contrib/adminpack's pg_file_unlink(), or an equivalent
> one-liner in plperlu or plpythonu. I don't know what the sqshf
> incantation does, though.
>
> This wouldn't be about fixing it so much as narrowing down where the
> problem is.

Will do. Before you spend a lot of time, let me get full confirmation
that this is the source of the issue and I'll bake a reproduction
script (there is still a possibility that something *else* caused it
and by happenstance we caught it here).

For posterity, sqshf does:

CREATE OR REPLACE FUNCTION sqshf(
ConnectionString TEXT,
FileName TEXT) RETURNS TEXT AS
$$#!/bin/bash

cat \
$2 \
| eval "sqsh $1 -L'datetime=%Y-%m-%d %H:%M:%S.%u' -G 7.0"

echo "Success"

$$ LANGUAGE plsh;

sqsh being the freetds wrapping linux console to sql server. So
basically not much.

shexec does:

CREATE OR REPLACE FUNCTION shexec(_cmd TEXT) RETURNS TEXT AS
$$#!/bin/bash

eval $1

$$ LANGUAGE plsh;

:-)

merlin