I think that when the query planner chooses to use an index to satisfy a
range scan, it verifies how the order of index keys is correlated with
physical order of rows in the table. If correlation is low, than the planner
wants to make sure that it reads each page only once.
So, first PG scans the
Thank you Tom, that's exactly what I meant. I believe the term "clustering
factor" is well-known:
http://use-the-index-luke.com/glossary/index-clustering-factor
What is the right word to use instead in Postgres universe?
--
View this message in context:
http://postgresql.nabble.com/Is-cluste
AlexK987 writes:
> I believe that tables' clustering factors are stored in pg_stats.correlation.
Dunno what you mean by "clustering factor", but pg_stats.correlation
stores the correlation coefficient between the sequence of the column's
data values and the sequence of their physical order. Corr
I believe that tables' clustering factors are stored in pg_stats.correlation.
Here is how I came to this conclusion:
create table narrow_table as
with numbers as(
select generate_series as n from generate_series(0,1048575))
select n as seq_number,
(n/1024) + ((n%1024)<<10) as rand_number
from nu
This worked, thank you!
--
View this message in context:
http://postgresql.nabble.com/How-to-connect-to-postgres-app-on-Mac-from-PgAdmin-tp5831604p5831609.html
Sent from the PostgreSQL - general mailing list archive at Nabble.com.
--
Sent via pgsql-general mailing list (pgsql-general@postgre
On 12/20/2014 12:46 PM, AlexK987 wrote:
I've installed postgres.app on my Mac. It did not prompt me for any password.
I am trying to connect to it via PgAdmin, from the same Mac laptop,
providing localhost as hostname, postgres as username, and no password.
I am getting "FATAL: role postgres doe
I've installed postgres.app on my Mac. It did not prompt me for any password.
I am trying to connect to it via PgAdmin, from the same Mac laptop,
providing localhost as hostname, postgres as username, and no password.
I am getting "FATAL: role postgres does not exist".
Which credentials should
Andomar writes:
> We run a Postgres 9.1 system that handles a lot of identical queries
> (around 500 per second.) The queries originate from a Python WSGI script
> running under Apache.
> Performance is an issue, and we suspect query planning might be an
> underlying cause. I've rewritten the
Hi
I'm working on a web credit card payment solution, so in a PCI-DSS environment,
and the auditors gave me trouble with one specific audit point for the
PostgreSQL database. They require the list of users in the database that had no
activity in the past 90 days to be deleted.
So far, it seems the
On 12/20/2014 07:22 AM, Pierre Ducroquet wrote:
On Saturday, December 20, 2014 06:57:54 AM Adrian Klaver wrote:
On 12/20/2014 06:40 AM, Pierre Ducroquet wrote:
Hi
I'm working on a web credit card payment solution, so in a PCI-DSS
environment, and the auditors gave me trouble with one specific
On Saturday, December 20, 2014 06:57:54 AM Adrian Klaver wrote:
> On 12/20/2014 06:40 AM, Pierre Ducroquet wrote:
> > Hi
> >
> > I'm working on a web credit card payment solution, so in a PCI-DSS
> > environment, and the auditors gave me trouble with one specific audit point
> > for the PostgreSQL
On 12/20/2014 06:40 AM, Pierre Ducroquet wrote:
Hi
I'm working on a web credit card payment solution, so in a PCI-DSS environment,
and the auditors gave me trouble with one specific audit point for the
PostgreSQL database. They require the list of users in the database that had no
activity in th
Hi
I'm working on a web credit card payment solution, so in a PCI-DSS environment,
and the auditors gave me trouble with one specific audit point for the
PostgreSQL database. They require the list of users in the database that had no
activity in the past 90 days to be deleted.
So far, it seems
Hi,
We run a Postgres 9.1 system that handles a lot of identical queries
(around 500 per second.) The queries originate from a Python WSGI script
running under Apache.
Performance is an issue, and we suspect query planning might be an
underlying cause. I've rewritten the scripts from ad-hoc
The way I would do it is to setup pgBouncer and control the user access
from there. Not to mention the connection pooling gains you would get from
pgBouncer itself. The simples way is to setup all of the users to connect
through pgBouncer, then the admin users connect to the Postgresql port
directl
David G Johnston wrote
> On Saturday, December 20, 2014, Mike Pultz [via PostgreSQL] <
> ml-node+s1045698n5831551h5@.nabble
>> wrote:
>
>> Hey David,
>>
>> > What version are you using?
>>
>> Sorry, I'm on 9.3.4.
>>
>> >Now() is a volatile function so the planner cannot omit partitions.
>> >
>>
16 matches
Mail list logo