On Thu, Sep 16, 2010 at 4:57 PM, Peter Eisentraut <peter_e@gmx.net> wrote:
> On tor, 2010-09-16 at 15:56 -0400, Robert Haas wrote:
>> I reiterate my earlier criticism of this whole approach: it seems to
>> assume that computing query progress is something inexpensive enough
>> that we can afford to do it regardless of whether anyone is looking.
>
> That assumption appears to hold so far.
It seems unlikely to hold in the general case, though, particularly if
you want to do it to be accurate. The problems with database-wide
VACUUM seem likely to be the tip of the iceberg.
> Anyway, do you have an alternative suggestion?
I think that there should be a function which returns just this one
piece of data and is not automatically called as part of select * from
pg_stat_activity. Then we could eventually decide to give backends a
way to know if that function had been invoked on them and how
recently.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise Postgres Company