Re: display previous query string of idle-in-transaction

(Sorry for delay..)

Guillaume Smet wrote:
> Being able to detect which application is running which query on the
> very same database with the very same user seems like something not so
> obvious and the use case seems to be pretty narrow. And IMHO, even if
> we suppose you can make the difference between the applications with
> only one query, you won't be able to limit your investigation to this
> application.
Yes, I won't be able to *completely* detect which application is running
long transaction with a last query.

But, as I said, I can get a hint for guessing causes from it.
And, as Simon said, I can detect a problem point with collaboration
from other information (app's log, app's source, operation procedure, and so on).

> So, in fact, I'm still not convinced. Could you detail a bit more how
> you plan to use it?
Well, Now, I can't get enough information from pg_stat_activity.
So, I have to check huge logs or reproduce the same problem.
(They are annoying works.)

If I can check last and more queries, I can use it as a hint for narrowing
down problem points with app's log and so on.
# And search the point and fix (or suggesting action) it.

I hope it would be able to narrowing down problem points more easily.

Best regards,
--
Tatsuhito Kasahara
kasahara(dot)tatsuhito(at)oss(dot)ntt(dot)co(dot)jp