QUERY PLANNER - Indexe mono column VS composite Index - Mailing list pgsql-performance

From Nicolas Paris
Subject QUERY PLANNER - Indexe mono column VS composite Index
Date
Msg-id CA+ssMOSn239yo+iJeXRLnsPEmnb-Ai-pQpQrp0=+Yt8wy6WHXA@mail.gmail.com
Whole thread Raw
Responses Re: QUERY PLANNER - Indexe mono column VS composite Index
List pgsql-performance
Hello,

My 9.4 database is used as datawharehouse. I can't change the queries generated.

first index  : INDEX COL (A,B,C,D,E)


In case of query based on COL A,  the query planner sometimes go to a seq scan instead of using the first composite index.

The solution is to add a second indexe (redondant)
second index : INDEX COL (A)

In case of query based on COL A, B, C, D, (without E) as well, it doesn't uses the first index and prefers a seq scan.

I could create a third indexe :
first index  : INDEX COL (A,B,C,D) 

But I hope there is an other solution for that (table is huge).

It seems that the malus for using composite indexes is high.

Question is : is there a way to make the composite index more attractive to query planner ? (idealy equivalent to mono column indexe)


Thanks by advance

pgsql-performance by date:

Previous
From: Jeff Janes
Date:
Subject: Re: pg_stat_all_indexes understand
Next
From: Guillaume Lelarge
Date:
Subject: Re: QUERY PLANNER - Indexe mono column VS composite Index