On Thu, Jan 28, 2021 at 9:30 AM Robert Haas wrote:
> On Thu, Jan 14, 2021 at 8:09 PM Peter Geoghegan wrote:
> > So dirty pages are debt that VACUUM can easily create, whereas buffer
> > misses are paid directly by VACUUM. It is its own backpressure, for
> > the most part. Making the costing stuff
On Thu, Jan 14, 2021 at 8:09 PM Peter Geoghegan wrote:
> So dirty pages are debt that VACUUM can easily create, whereas buffer
> misses are paid directly by VACUUM. It is its own backpressure, for
> the most part. Making the costing stuff highly sensitive to dirtying
> pages (but not sensitive to
On Thu, Jan 21, 2021 at 5:12 PM Peter Geoghegan wrote:
> I'm going to go ahead with committing my patch to lower the default
> next week. If anybody has any objections to that plan, please speak
> up.
Just pushed a commit that reduced the default for vacuum_cost_page_miss to 2.
One more thing on
On Thu, Jan 14, 2021 at 8:34 PM Masahiko Sawada wrote:
> +1 for this change. Lowering to 2 also looks good to me.
I'm going to go ahead with committing my patch to lower the default
next week. If anybody has any objections to that plan, please speak
up.
It doesn't really need to be said again, b
On Thu, Jan 14, 2021 at 9:24 AM Peter Geoghegan wrote:
>
> On Wed, Jan 6, 2021 at 7:43 PM Peter Geoghegan wrote:
> > More concretely, we could perhaps lower vacuum_cost_page_miss to 5. It
> > has had the value as 10 as its default since 2004 (just like
> > vacuum_cost_page_dirty, whose default ha
On Thu, Jan 14, 2021 at 10:42 AM Robert Haas wrote:
> There are also users I've seen get hosed by vacuuming too
> aggressively. I have seen this happen in two ways. One is too much
> dirty data. The other is too much read I/O, pushing hot data out of
> the cache, leading to a storm of random I/O l
On Thu, Jan 14, 2021 at 9:29 AM Magnus Hagander wrote:
> Do you have any actual metrics between specifically choosing the value
> 3? Or is that off a gut feeling?
I have no metrics, exactly, but I'm sure that the trend I mentioned
about page cleaning/dirtying being the bottleneck more and more th
On Thu, Jan 14, 2021 at 12:29 PM Magnus Hagander wrote:
> +1 for this in principle.
I'm not opposed to this change and I agree that the relative expense
of dirtying a page is higher than what the current defaults suggest.
So I also think Peter is going in the right general direction, though
like
On Thu, Jan 14, 2021 at 1:24 AM Peter Geoghegan wrote:
>
> On Wed, Jan 6, 2021 at 7:43 PM Peter Geoghegan wrote:
> > More concretely, we could perhaps lower vacuum_cost_page_miss to 5. It
> > has had the value as 10 as its default since 2004 (just like
> > vacuum_cost_page_dirty, whose default ha
On Wed, Jan 6, 2021 at 7:43 PM Peter Geoghegan wrote:
> More concretely, we could perhaps lower vacuum_cost_page_miss to 5. It
> has had the value as 10 as its default since 2004 (just like
> vacuum_cost_page_dirty, whose default has also not been changed since
> the start). These defaults were de
On Wed, Jan 6, 2021 at 5:39 PM Masahiko Sawada wrote:
>
> On Mon, Dec 28, 2020 at 5:17 AM Peter Geoghegan wrote:
> >
> > Simply decreasing vacuum_cost_page_dirty seems like a low risk way of
> > making the VACUUM costing more useful within autovacuum workers.
> > Halving vacuum_cost_page_dirty to
On Wed, Jan 6, 2021 at 5:39 PM Masahiko Sawada wrote:
> Perhaps you meant to decrease vacuumm_cost_page_miss instead of
> vacuum_cost_page_dirty?
You're right. Evidently I didn't write this email very carefully.
Sorry about that.
To say it again: I think that a miss (without dirtying the page)
s
On Mon, Dec 28, 2020 at 5:17 AM Peter Geoghegan wrote:
>
> Simply decreasing vacuum_cost_page_dirty seems like a low risk way of
> making the VACUUM costing more useful within autovacuum workers.
> Halving vacuum_cost_page_dirty to 5 would be a good start, though I
> think that a value as low as 2
vacuum_cost_page_miss has a default value of 10, while
vacuum_cost_page_dirty has a default value of 20. This has been the
case since cost-based delays were introduced by commit f425b605f4e
back in 2004. The obvious implication is that dirtying a page is on
average only twice as expensive as a sing
14 matches
Mail list logo
