> On Aug. 25, 2016, 12:30 a.m., Navina Ramesh wrote: > > docs/learn/documentation/versioned/api/overview.md, line 22 > > <https://reviews.apache.org/r/50174/diff/2/?file=1455859#file1455859line22> > > > > "You should implement StreamTask for synchronous process, e.g. a > > computation that does not involve remote calls" > > > > This is not very clear. To me, it sounds like we are recommending a > > task interface based on a use-case. Instead, I think it will be useful to > > explain what you mean by "synchronous processing of a stream message" and > > compare it with asynchronous processing. This can be followed by the remote > > call example.
Added the definition there. > On Aug. 25, 2016, 12:30 a.m., Navina Ramesh wrote: > > docs/learn/documentation/versioned/api/overview.md, line 54 > > <https://reviews.apache.org/r/50174/diff/2/?file=1455859#file1455859line54> > > > > You have mentioned "processAsync". Your code sample above show > > "process". Please fix it. Thanks for catching it. > On Aug. 25, 2016, 12:30 a.m., Navina Ramesh wrote: > > docs/learn/documentation/versioned/container/event-loop.md, line 24 > > <https://reviews.apache.org/r/50174/diff/2/?file=1455860#file1455860line24> > > > > There is no sharing of task state in Samza. Each task guarantee > > isolation from the others. What exactly are you referring to here? Unless I > > have misunderstood the semantics provided by processing in multiple threads The shared state refers to global in-memory states, like singletons and static vars. > On Aug. 25, 2016, 12:30 a.m., Navina Ramesh wrote: > > docs/learn/documentation/versioned/container/event-loop.md, line 43 > > <https://reviews.apache.org/r/50174/diff/2/?file=1455860#file1455860line43> > > > > I think you are trying to explain 2 things: > > 1. semantics of event loop when using AsyncStreamTask > > 2. Semantics of event loop when using StreamTask with threadPool size > > > 1 > > > > Can you try to elaborate more on this? Perhaps just separating > > semantics of event loop for StreamTask and AsyncStreamTask will add more > > clarity. > > > > Also, the processing order guarantees need to be called out. It is > > possible to process out-of-order within a partition if the task threadpool > > size is >1. This is a very important behavior change that needs to be > > documented here. I ended up by adding section named "Semantics for Synchronous Tasks v.s. Asynchronous Tasks". Please take a look again. > On Aug. 25, 2016, 12:30 a.m., Navina Ramesh wrote: > > docs/learn/documentation/versioned/container/event-loop.md, line 47 > > <https://reviews.apache.org/r/50174/diff/2/?file=1455860#file1455860line47> > > > > nit: "through the standard InitableTask, ClosableTask, StreamTask / > > AsyncStreamTask, and WindowTask." fixed. > On Aug. 25, 2016, 12:30 a.m., Navina Ramesh wrote: > > docs/learn/documentation/versioned/container/event-loop.md, line 49 > > <https://reviews.apache.org/r/50174/diff/2/?file=1455860#file1455860line49> > > > > Does anything change with respect to adding "task.subtask.class" in the > > process() method of AsyncStreamTask? Please check if this does not change > > the expected behavior in SAMZA-437. We use that at LinkedIn for > > restliWrapper (I think). This is actually not implemented in open source, only in LinkedIn. > On Aug. 25, 2016, 12:30 a.m., Navina Ramesh wrote: > > docs/learn/documentation/versioned/jobs/configuration-table.html, line 360 > > <https://reviews.apache.org/r/50174/diff/2/?file=1455861#file1455861line360> > > > > Isn't there a performance impact if we use AsyncStreamTask by default? > > Esp. on jobs that don't need any async processing. > > Why is the default value false? > > > > If users upgrade to the new version and automatically start using > > multithreaded execution, will they see any performance impact? The default of new AsyncRunLoop is still running in a single thread, and the performance is on par with the previous RUnLoop. So the users shouldn't see any perf difference. - Xinyu ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/50174/#review146554 ----------------------------------------------------------- On Aug. 30, 2016, 12:49 a.m., Xinyu Liu wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/50174/ > ----------------------------------------------------------- > > (Updated Aug. 30, 2016, 12:49 a.m.) > > > Review request for samza, Chris Pettitt, Navina Ramesh, and Yi Pan (Data > Infrastructure). > > > Repository: samza > > > Description > ------- > > Update samza web docs with new multithreading api, core and configs. > > > Diffs > ----- > > docs/learn/documentation/versioned/api/overview.md > 6712344e84e19883b857e00549db2acb101c7e0e > docs/learn/documentation/versioned/container/event-loop.md > 116238312df7071747cbbc14bc9c46f558755195 > docs/learn/documentation/versioned/jobs/configuration-table.html > 54c52981c3055b398ee60af50eeaf2592ed0e64f > > Diff: https://reviews.apache.org/r/50174/diff/ > > > Testing > ------- > > Test the web pages locally. > > > Thanks, > > Xinyu Liu > >
