Question

Let me get to my example:

For the ID=>values 0=>87, 1=>24, 2=>82, 3=>123, 4=>34, 5=>61,
increment all values for keys between 1 and 4 by 10

For a multi-row operation like this, does Riak offer atomicity; ie this operation either fails or succeeds, without dirtying the data partially?

Do queries aggregating on the rows when they are updating see consistent results?

---

I saw no place which dealt with this question explicitly. But I guess the “tunable CAP” controls set to “enable consistency and partition tolerance” seems like the key.

Answer 1

No.

Riak has no concept of atomicity overall (it’s an eventually consistent system), and also does not have any concept of a “transaction” where multiple K/V pairs can be modified or read as a set.