redshift performance effect if use 32 character long unique string as Primary key instead of auto incremented int

In my analytics application I am creating 32 char long unique string for each hit as user_session.

In current implementation, I am storing user_session in tbl_sesssion in session column and I am using auto incremented session_id as primary key.

Problem is for each new hit I need to perform select * from tbl_session where session = 'user_session', before inserting in tbl_session(as I want unique record for session) which is costly. I think this can be solved if I use user_session directly as primary key, as in this case select operations will not be required, only insert operations will happen.

So if I use user_session directly as primary key is it will cause performance degradation?

I am using amazon redshift for storing data, and looking at 1 million hits per day.