The restore side speedups suggested by Yang seem reasonable and can potentially speed up the process. We can even go a bit further by starting the new postgres in a --binary-upgrade mode and skip some of these locks completely.
On Mon, Jul 08, 2024 at 03:22:36PM +0800, 杨伯宇(长堂) wrote: > Besides, https://commitfest.postgresql.org/48/4995/ seems insufficient to > this situation. Some time-consuming functions like check_for_data_types_usage > are not yet able to run in parallel. But these patches could be a great > starting point for a more efficient parallelism implementation. Maybe we can > do it later.
I actually just wrote up the first version of the patch for parallelizing the data type checks over the weekend. I'll post it shortly.
