Re: [PERFORM] Poor Performance on Postgres 8.0

[Pallav Kalva]

On 7.4 I get

Advertising







select * from pg_stats
where tablename = 'attribute' and attname = 'fknamestringid';
schemaname | tablename |    attname     | null_frac | avg_width | 
n_distinct |                  most_common_vals                   
|                                  
most_common_freqs                                  |                     
histogram_bounds                     | correlation
------------+-----------+----------------+-----------+-----------+------------+-----------------------------------------------------+-------------------------------------------------------------------------------------+----------------------------------------------------------+-------------
common     | attribute | fknamestringid |         0 |         4 
|        124 | {2524,2434,2523,2599,2595,2592,2596,2528,2586,2446} | 
{0.132333,0.13,0.0766667,0.0373333,0.0366667,0.0333333,0.031,0.029,0.0263333,0.019} 
| {2433,2441,2455,2462,2473,2479,2484,2492,2505,2574,2598} |    -0.22864
(1 row)

On 8
select * from pg_stats
where tablename = 'attribute' and attname = 'fknamestringid';
schemaname | tablename |    attname     | null_frac | avg_width | 
n_distinct |                  most_common_vals                   
|                                 
most_common_freqs                                 |                     
histogram_bounds                     | correlation
------------+-----------+----------------+-----------+-----------+------------+-----------------------------------------------------+-----------------------------------------------------------------------------------+----------------------------------------------------------+-------------
common     | attribute | fknamestringid |         0 |         4 
|         80 | {2524,2434,2530,2522,2525,2523,2527,2526,2574,2531} | 
{0.219333,0.199333,0.076,0.0643333,0.0616667,0.05,0.0453333,0.042,0.04,0.0286667} 
| {2437,2528,2529,2538,2539,2540,2554,2562,2575,2584,2637} |   0.0274016

Tom Lane wrote:
I was able to duplicate this behavior with dummy data that had only a
few distinct values for fknamestringid --- the planner then thinks that
the index probe into attribute will match a lot of rows and hence take a
long time.  Could we see your pg_stats row for fknamestringid, ie
select * from pg_stats
where tablename = 'attribute' and attname = 'fknamestringid';
It would be interesting to see the same for your 7.4 installation too.
                        regards, tom lane
 


---------------------------(end of broadcast)---------------------------
TIP 7: don't forget to increase your free space map settings