On Thu, Feb 16, 2017 at 6:17 AM, Simon Riggs <simon@2ndquadrant.com> wrote:
> On 15 February 2017 at 08:07, Masahiko Sawada <sawada.mshk@gmail.com> wrote:
>> It's a bug. Attached latest version patch, which passed make check.
>
> In its current form, I'm not sure this is a good idea. Problems...
>
> 1. I'm pretty sure the world doesn't need another VACUUM parameter
>
> I suggest that we use the existing vacuum scale factor/4 to reflect
> that indexes are more sensitive to bloat.
I do not think it's a good idea to control multiple behaviors with a
single GUC. We don't really know that dividing by 4 will be right for
everyone, or even for most people. It's better to have another
parameter with a sensible default than to hardcode a ratio that might
work out poorly for some people.
> 2. The current btree vacuum code requires 2 vacuums to fully reuse
> half-dead pages. So skipping an index vacuum might mean that second
> index scan never happens at all, which would be bad.
Maybe. If there are a tiny number of those half-dead pages in a huge
index, it probably doesn't matter. Also, I don't think it would never
happen, unless the table just never gets any more updates or deletes -
but that case could also happen today. It's just a matter of
happening less frequently.
I guess the question is whether the accumulation of half-dead pages in
the index could become a problem before the unsetting of all-visible
bits in the heap becomes a problem. If the second one always happen
first, then we don't have an issue here, but if it's possible for the
first one to become a big problem before the second one gets to be a
serious issue, then we need something more sophisticated.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company