Re: Request: pg_cancel_backend variant that handles 'idle in transaction' sessions

On Thu, Nov 5, 2015 at 12:31 PM, Joe Conway <mail@joeconway.com> wrote:
> On 11/05/2015 10:09 AM, Pavel Stehule wrote:
>> On 5.11.2015 19:02 Merlin Moncure wrote:
>>> Thus, I think we have consensus that transaction_timeout is good -- it
>>> would deprecate statement_timeout essentially.  Likewise,
>>> pg_cancel_transaction is good and would deprecate pg_cancel_backend;
>>> it's hard for me to imagine a scenario where a user would call
>>> pg_cancel_backend if pg_cancel_transaction were to be available.
>>
>> I am sorry, I see a consensus between you and Stephen only.
>
> S
> t                    C
> a<-------------<transaction>--------------->E
> r    A      B     A      B      A           n
> t <idle> <stmt> <idle> <stmt> <idle>        d
> |--------======--------======---------------|
>
> Currently we can set timeout and cancel for period B (<stmt>). I can see
> based on this discussion that there are legitimate use cases for wanting
> timeout and cancel for any of the periods A, B, or C.
>
> I guess the question then becomes how we provide that coverage. I think
> for coverage of timeout you need three individual timeout settings.
> However for cancel, it would seem that pg_cancel_transaction would cover
> all three cases.

Agreed on all points.

Tom noted earlier some caveats with the 'idle' timeout in terms of
implementation.  Maybe that needs to be zeroed in on.

merlin