Aurora cluster writer 2xlarge with 20GB memory.
I was doing bunch of updates on all orders, members, Catalogs table. The scope was to replace some information for certain fields.
The operation was quite heavy because the tables have more than 50 Milions rows and it had a condition with like and percent at the beginning.
It didn't generate the best execute plan because that condition doesn't use the index on that field.
I had to develop a store procedure that had in input the lower and upper id limits, the primary key, and the range that it needs to add at every run.
Quite simple procedure that allow to split up the updates in smaller pieces.