transform and synchronize data across multiple systems - mysql

Question

I'll quickly explain the Data Flow Diagram above.
The main process in the bottom left corner and the mongodb datastore next to it are two main components of my system. Simply put, main process is gathering data from a MySQL system which serves as a datastore for other backend systems in our company. Other systems which are external to my system are constantly changing data in their respective MySQL DB. The main process is transforming data from those systems, not actually changing the original schema but adding more information to it and sometimes updating its values, NOT SCHEMA. The transformed data is used by our mobile apps ie external entity next to mongodb datastore in DFD. Now, everything works fine when I use the system to create a new copy of transformed data, at that moment it is synchronized with all other systems in terms of data.
The problem is,
When I try to further transform the data later at some point, I want to be able to notify user of changes and synchronize it(if the user wants to) with original data as other external systems or even my own process could have updated the data.
{
"data_gathered_by_process": [
{
"id": "DB1",
"original data field 1": "original value 1"
},
{
"id": "DB2",
"original data field 2": "original value 2"
},
{
"id": "DB3",
"original data field 3": "original value 3"
}
]
}
This could be transformed into
{
"transformed data": [
{
"id": "DB1",
"original data field 1": "transformed value 1",
"additional field added by process": "value"
},
{
"id": "DB2",
"original data field 2": "original value 2",
"additional values": ["one", "two"]
},
{
"id": "DB3",
"original data field 3": "original value 3"
}
]
}
Now the original data could again be changed this way
{
"data_gathered_by_process": [
{
"id": "DB1",
"original data field 1": "changed to some other value"
},
{
"id": "DB2",
"original data field 2": "original value 2"
},
{
"id": "DB3",
"original data field 3": "original value 3"
}
]
}
I'm thinking of implementing something like this:
add a last_updated timestamp on entities of DB1, DB2, DB3 and also store it in transformed copies of data. When working on already transformed data, check timestamps of all entities one by one and update if mismatched. I'll first notify user that original data has changed since and if he wants to make changes use the same logic. But this would be a processing overhead as there are more than ten entities with each having different set of properties.
I think there's a better way to do this, if only someone could point me in the right direction.

Answer 1

(From a MySQL perspective...)
A single dataset (all the DBs), living on 3 machines. Those 3 machines in some kind of Replication.
If it is specifically "3", then Galera Cluster is an excellent choice. Each client can write to its nearby Galera "node"; that data will be immediately replicated to the other two nodes.
If the "3" is likely to grow over time, I would have 1 "Primary" with some number of Replicas. This topology requires that all writes go to the Primary. How far (ping time) is it from each client to where the Primary might live?
The number of Replicas could be zero -- everything is done on the Primary. Or it could be a growing number of servers; this is useful if the read traffic is too high to handle in a single server.
Both approaches (Galera vs Primary+Replicas) force every write to go to all servers, thereby eliminating the synchronization you describe.
(A 3rd approach is "InnoDB Cluster".)