Let's say I have a DDB table similar to the following:
LocId (hash)
Index (range)
...
The data for a single "Loc" would be spread out across multiple records (I'm not looking for advice on how to restructure the data, I have reasons for this approach). If we have a DDB stream enabled and wired to a Lambda function will I be guaranteed to get all changes to any LocId record on the same shard?
I don't care too much about FIFO, I just want to make sure that if, say, fifteen different indexes were updated for the same LocId at one time, the changes wouldn't be spread out over multiple instances of the Lambda function.
Is it correct that your use case is not guaranteed? The blog states that only the order of a primary key (Hash+Range) will be in at-most one shard.