On Thu, Aug 23, 2018 at 6:35 AM Ivan Kelly <iv...@apache.org> wrote: > >> We currently create an OrderedExecutor and an OrderedScheduler in the > >> client. An OrderedScheduler is an OrderedExecutor. Moreover, it's very > >> seldom used (basically for polling LAC, speculative reads and explicit > >> flush. > > > > Why do they exist? Are they only legacy from past or is there any > specific > > reason? > > Accident I think.
I don't think it is accidently. OrderedExecutor has performance advantages than OrderedScheduler. A bit background on this: - OrderedScheduler was introduced by me. And I changed existing OrderedSafeExecutor to be extending from OrderedScheduler. Trying to standarize to one `OrderedScheduler`. - I think Matteo noticed performance difference between `executor` and `scheduler`, so he made the change https://github.com/apache/bookkeeper/commit/46171e67e526702487641438144f28b7eb1aa07b . So the `executor` is used as the main callback executor, since it just requires ordering but doesn't need scheduling capability. the `scheduler` is used for scheduling tasks but doesn't require ordering. > Previously, OrderedExecutor didn't implement > scheduling, and BookKeeper had a ScheduledExecutorService member. > This changed with > > https://github.com/apache/bookkeeper/commit/46171e67e526702487641438144f28b7eb1aa07b > . > > > Is there any per ledger ordering requirement that we will break (or cause > > deadlock) if we have a single queue per ledger? > > Any ordering that depends on this is broken. No operations should > block on either of these executors, and I don't think any do. > > -Ivan >
