Should be handled in RPC queues first, before mutation op reaches RingBuffer, I think. But to answer your question: the only guarantee HBase provide (promises) is strictly consistent writes and atomicity for a single row mutation. No, won't break anything. Permutations of mutations in an execution pipeline is safe, because they can't violate HBase promises.
-Vlad On Wed, Nov 7, 2018 at 1:49 PM Jing Liu <jingliu...@gmail.com> wrote: > Hi, > > I'am trying to add priority to schedule different types of requests in > HBase. But the Write-ahead logging use RingBuffer which > is essentially a FIFO queue makes it hard. In this case, let's say if the > low priority request already queued in the RingBuffer, the high priority > request can not be executed before all those queued low priority request. > I'm wondering if I change the FIFO queue into Priority-based queue > will violate the write consistency guarantee or other issues? > > Thanks, > Jing >