The workaround means doing two ALTERs which might not be feasible in certain situations with big tables as it will increase the alter time to twice of what it should be. Since the old PK column will essentially be unique and each record will have a unique value for each row that comes into the new table till pt-osc is running, so I do not see how this can be an issue, or may be I am missing something :)
Hi Daniel,
The workaround means doing two ALTERs which might not be feasible in certain situations with big tables as it will increase the alter time to twice of what it should be. Since the old PK column will essentially be unique and each record will have a unique value for each row that comes into the new table till pt-osc is running, so I do not see how this can be an issue, or may be I am missing something :)