I have following index question:
First a have this table definition:
mytable
- id (pk)
- column1 (fk to any other table)
- column2
index 1: id
index 2: column1
…when a call this query, then it is not very fast because only one column index is used
select * from mytable where column1 = ?? order by column2
…then a changed mytable to this:
mytable
- id (pk)
- column1 (fk)
- column2
index 1: id
index 2: column1, column2
in this table I created index with both columns and droped column1 index
and now my question: is there any other performance problem when a do this?