Re: implementing query timeout

Jessica Perry Hekman wrote:
> I am still unable to post to the hackers list. I sent a second attempt, a
> few days after subscribing; I sent mail to pgsql-hackers-owner (no
> response, though maybe that didn't go through either); I subscribed from a
> second account, waited a few hours, and send mail (twice) from there
> (again, no bounce and it didn't go through). I am at a loss. Is there some
> special way to subscribe? (I filled my address into the form at
> http://webmail.postgresql.org/cgi-bin/mj_wwwusr?domain=postgresql.org&func=lists-long-full&extra=pgsql-hackers,
> but maybe that just lets me read and not post?).
>
> I'm sorry to keep asking for your help on this, but I am frustrated and
> not sure what else to do at this point.

Jessica, you had asked about implementing a jdbc timeout a while back,
and we got stuck on implementing it in the backend with an alarm()
timeout and a Cancel, and the fact that SET in a transaction would not
roll back.

Well, we have committed fixes to the current CVS to roll back a SET in a
transaction:

    test=> SHOW australian_timezones;
    INFO:  australian_timezones is off
    test=> BEGIN;
    BEGIN
    test=> SET australian_timezones TO true;
    SET
    test=> SHOW australian_timezones;
    INFO:  australian_timezones is on
    test=> ABORT;
    ROLLBACK
    test=> SHOW australian_timezones;
    INFO:  australian_timezones is off

This means we can now create a SET variable that does an alarm/cancel
timeout in a transaction, and know that it can safely bracket a
statement because it will reset if the transaction aborts.

Are you still interested in working on this?

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 853-3000
  +  If your life is a hard drive,     |  830 Blythe Avenue
  +  Christ can be your backup.        |  Drexel Hill, Pennsylvania 19026