Re: Indexes not always used after inserts/updates/vacuum analyze

[Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>]

I said:
>> symbol_data | symbol_name |         0 |         7 |     152988 | 
>> {EBALX,ELTE,LIT,OEX,RESC,BS,ESH,HOC,IBC,IDA} | 
>> {0.0183333,0.0173333,0.00166667,0.00166667,0.00166667,0.00133333,0.00133333,0.00133333,0.00133333,0.00133333} 
>> | {A,BMO,DBD,FSCHX,IIX,MAS,NSANY,PTEC,SR,UTIL,_^^VPM} |    0.128921
>> (1 row)

> What this says is that in the last ANALYZE, EBALX accounted for 18% of
> the sample, and ELTE for 17%. 

Argh, make that 1.8% and 1.7%.

That's still orders of magnitude away from what you say the correct
frequency is, however: 687 out of 20+ million.  I'd like to think that
the statistical sampling would be unlikely to make such a large error.

			regards, tom lane