Although from what you have mentioned recreating the table with a new PI ( mind it not PK they are not the same in Teradata) should be a viable option, but I would suggest you do the below :
a) Do a data demographics analysis. You can use the Help Stats functionality. b) Based on the above see if the skew is not because of lots of NULLs in the current PI. I had a similar experience when a table was very much skewed, The problem was a large number of Nulls in the PI columns. This was resulting in a Hot AMP. I filled the Nulls with Dummy value and that promtly removed the skew. c) Use Index Wizard if you want.
select hashamp(hashbucket(hashrow(job_id))) ,count(*) from log_table group by 1;
11 14 0 32 7 4 13 13122 this is the problem 1 30 5 32 9 236 4 6 2 10 12 12 3 883 10 8 8 30 6 28
I know exactly why this happened. There's a very valid if very stupid explanation. There is a date field i could use to give a very even partition. There is also one other column I could use(a description column).
What I really want to know is: If i drop the table and re-create with a new PI will it automatically rebalance the data across the amps?