Re: pg_terminate_backend idea

On 2005-06-22, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Andrew - Supernews <andrew+nonews(at)supernews(dot)com> writes:
>> On 2005-06-22, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>>> Andreas Pflug <pgadmin(at)pse-consulting(dot)de> writes:
>>>> I've seen cancel *not* working.
>>>
>>> Even a moment's perusal of the code will prove that there is no
>>> situation in which a backend will respond to SIGTERM but not SIGINT
>
>> "idle in transaction". (or "idle" for that matter, but that's usually less
>> significant.)
>
> In that case there's no query to cancel, so I would dispute the claim
> that that constitutes "not working".

You are totally missing the point.

A backend that is "idle in transaction" is holding locks and an open xid
that cannot be cleared by anything short of SIGTERM.

Whether the fact that it ignores SIGINT is intentional or not is irrelevent,
the fact is that this is the classic scenario where SIGTERM is effective and
SIGINT is not.

--
Andrew, Supernews
http://www.supernews.com - individual and corporate NNTP services