On Thu, Nov 3, 2022 at 12:54 AM Michal Hocko <mhocko@xxxxxxxx> wrote: > > On Wed 02-11-22 13:08:08, Yang Shi wrote: > [...] > > So I think we narrowed down to two options: > > 1. Preserve the interleave behavior but bail out if the target node is > > not online (it is also racy, but doesn't hurt) > > I do not think there is merit in the interleave patch is dubious to say > the least. > > > 2. Remove the node balance code entirely > > Yes, removing the balancing makes sense but I would still hope that we > do not fail too easily if the range is populated on multiple nodes > equally. In practice it will likely not matter much I guess but setting > up all nodes with top score is just easy to achieve. OK, thanks. I will come up with a patch to allow fallback between the nodes, it should be largely based on the change suggested by you. > -- > Michal Hocko > SUSE Labs
