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

On 11/3/15 8:44 AM, Merlin Moncure wrote:
>> Actually, one other thing that would help is to have the ability to turn
>> >this into an ERROR:
>> >
>> >begin;
>> >WARNING:  there is already a transaction in progress
> curious: does the SQL standard define this behavior?
>
> Anyways, we've pretty studiously avoided (minus a couple of
> anachronisms) .conf setting thats control behavior of SQL commands in
> a non performance way.

If we had an event trigger on BEGIN and a way to tell whether we were 
already in a transaction this wouldn't need to be a config setting.

> IMO, this as yet another case for 'stored procedures' that can manage
> transaction state: you could rig up your own procedure: CALL
> begin_tx_safe(); which would test transaction state and fail if
> already in one.  This doesn't help you if you're not in direct control
> of application generated SQL but it's a start.  Barring that, at least

Even then it would be very easy to mess this up.

> warnings tend to stand out in the database log.

That depends greatly on how much other stuff is in the log. Something 
else I wish we had was the ability to send different log output to 
different places.
-- 
Jim Nasby, Data Architect, Blue Treble Consulting, Austin TX
Experts in Analytics, Data Architecture and PostgreSQL
Data in Trouble? Get it in Treble! http://BlueTreble.com