True to forgettable SQL-ish (did you know that indexes are not addressed by the SQL standard?) syntax, I got it slightly wrong:
$ psql
fdr=> \h CREATE INDEX
Command: CREATE INDEX
Description: define a new index
Syntax:
CREATE [ UNIQUE ] INDEX [ CONCURRENTLY ] [ name ] ON table [ USING method ]
( { column | ( expression ) } [ COLLATE collation ] [ opclass ] [ ASC | DESC ] [ NULLS { FIRST | LAST } ] [, ...] )
[ WITH ( storage_parameter = value [, ... ] ) ]
[ TABLESPACE tablespace ]
[ WHERE predicate ]
So, rather: CREATE INDEX CONCURRENTLY foo ....
I use this all the time, and am flabbergasted how people can do without it. I feel like migration frameworks should make it the default with Postgres.
It's too bad it can't be mixed with transactional DDL, but because indexes are not logical changes, I don't really care as much, even if it is dissatisfying.
So, all in all, for those who want to take advantage of this feature in Postgres:
Stop doing this:
CREATE INDEX foo ...
Start doing this:
CREATE INDEX foo CONCURRENTLY ...
For the cost of one keyword, your index additions can be a non-event.