On 3/6/19 8:04 PM, Robert Haas wrote:
> On Wed, Mar 6, 2019 at 1:53 PM Alvaro Herrera <alvherre@2ndquadrant.com> wrote:
>> On 2019-Feb-08, Tomas Vondra wrote:
>>> I'm wondering if we should just get rid of all such optimizations, and
>>> make the unowned list doubly-linked (WIP patch attached, needs fixing
>>> the comments etc.).
>>
>> +1 for that approach.
>
> +1 for me, too.
>
>> Did you consider using a dlist?
>
> Maybe that is worthwhile, but this is a smaller change, which I think
> should count for quite a bit. Nothing keeps somebody from doing the
> dlist change as a separate patch, if desired.
>
Yeah, although now that I think about it I wouldn't expect the dlist
version to be much more complicated. We access next_unowned_reln on two
or three places, IIRC, so switching to dlist would be trivial I think.
What worries me more is that I observe the opposite behavior than what's
described in comment for b4166911 (which is from 2018, so not that old)
and 279628a0a7 (from 2013). So what changed since then? Seems fishy ...
regards
--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services