Re: Clustered primary key question

[Doron Grinstein]

If I understand your suggestion correctly, you're saying to use an auto-incrementing INT as the primary key and a second field as a GUID. With an Index on the second field.

This approach does not work for multi master scenarios. I want the system to survive network partitions. Assuming I have 3 db hosts, a, b, c: I want users close to a to insert into a, those close to b to insert into b and those close to c to insert into c. Suppose users close to a can reach a, but a cannot reach b. Users of b can still reach b and insert there. When connectivity between a and b is re-established, replication will fail if an id in a is already taken by an insertion that took place in b.

Unless I misunderstood what you meant in your post or I misunderstand how replication works in MySql (each replica can have different primary key?) then your proposed solution does not seem to help.

I do appreciate your continued interest in this topic and honestly hope that there is a viable approach to using GUIDs as primary keys in MySql.