Consistency becomes an issue as soon as you have multiple copies of a data item, especially in a distributed system.
The CAP Theorem says that you can't have all three of Consistency, Availability and Partition Tolerance.
If you choose A and P, then you have to give up Consistency (to some extent). This means that different parts of your system may have different values for the data item. The different parts of the system will continually try to keep consistent with one another by communicating, but this may take a little time (hence "eventual" consistency).
Some mechanisms/protocols for achievinig consistency are described in the answers to this related question.
So talking about "how to get eventual consistency" is missing the point - eventual (or no) consistency is what you get by default in a distributed system, unless you are very careful (and sacrifice at least one of Availability and Partition Tolerance). Fully consistent systems have to work harder to ensure that all copies are correctly synchronized.