As of now, we use CSV lookups but some of the lookups are around 2 GB which is creating a problem in SH replication.
As far as I know, in case of any change in CSV lookup, a new CSV is created and this is considered as a new object and that's why this complete CSV gets replication to other SH. I mean to say diff is not replication in case of CSV lookups.
But in the KV store only transactions related to "changed rows" will be replicated so I think moving to KV store should solve the SH replication issue.
I know there might be an issue related to opslog size, but its limit is 1 GB and it would contain only transactions (for changed rows) so overall size will be much less compared to complete CSV lookup.
To be on the safe side I will increase its size to 1.5 GB.
So, moving to KV store will solve the SH replication problem until lots of transactions fill opslog file, which is very rare (given that only changes get replicated).
Am I right?
