On Thu, Nov 9, 2023 at 11:40 AM Kyotaro Horiguchi <horikyota....@gmail.com> wrote: > > At Thu, 9 Nov 2023 09:53:07 +0530, Amit Kapila <amit.kapil...@gmail.com> > wrote in > > Michael, Horiguchi-San, and others, do you have any thoughts on what > > is the best way to proceed? > > As I previously mentioned, I believe that if rejection is to be the > course of action, it would be best to proceed with it sooner rather > than later. On the other hand, I am concerned about the need for users > to perform extra steps depending on the source cluster > conrfiguration. Therefore, another possible approach could be to > simply ignore the given settings in the assignment hook rather than > rejecting by the check hook, and forcibuly apply -1. > > What do you think about this third approach? >
I have also proposed that as one of the alternatives but didn't get many votes. And, I think if the user is passing a special value of max_slot_wal_keep_size during the upgrade, it has to be a special case, and rejecting it upfront doesn't seem unreasonable to me. -- With Regards, Amit Kapila.
