The current set of watermark estimators in Apache Beam for UnboundedSource are: SQS - tracks the timestamp of the last unacked message (does not report monotonically increasing watermarks and assumes that the system will make sure to lower bound what is being reported) AMP - tracks timestamp of last message GCP Pubsub <https://github.com/apache/beam/blob/784d18b7ac89f87dd7fbf2861ee877f5b6070276/sdks/java/io/google-cloud-platform/src/main/java/org/apache/beam/sdk/io/gcp/pubsub/PubsubUnboundedSource.java#L950>: uses last minute of data to compute watermark but also handles cases where there is no data for long periods of time Kafka: Supports a watermark policy function <https://github.com/apache/beam/blob/c2f0d282337f3ae0196a7717712396a5a41fdde1/sdks/java/io/kafka/src/main/java/org/apache/beam/sdk/io/kafka/TimestampPolicy.java#L62> that tracks one of (processing time, ingestion time, or custom user based version that gets to see the individual records) Kinesis: Supports a watermark policy function <https://github.com/apache/beam/blob/c2f0d282337f3ae0196a7717712396a5a41fdde1/sdks/java/io/kinesis/src/main/java/org/apache/beam/sdk/io/kinesis/WatermarkPolicy.java> that tracks one of (processing time, arrival time, or custom user based version that gets to see the individual records)
Has anyone written or knows of custom user based watermark policy functions that have been implemented for Kafka or Kinesis? On Mon, Mar 9, 2020 at 9:54 AM Ismaël Mejía <ieme...@gmail.com> wrote: > Thanks for the explanation on Watch + FileIO it is really clear. Extra > question > related to WatermarkEstimator, is it supposed to be called in pipelines at > the > same exact moments that getWatermark is today for Unbounded sources? > Yes. DoFn authors will invoke it either manually after they tell us that the watermark has advanced or "automatically" via the timestamp observing watermark estimators. > (slightly unrelated) There is an open JIRA for an issue related to watch > and > watermark progress that we should probably investigate. I will add details > there > and ping you once I have the time to go back to the subject. > https://issues.apache.org/jira/browse/BEAM-9134 The Watch stuff is built on top of a very experimental version of SDF for classic runners. I know that some people have used it effectively though with watermark tracking working for them. > Thanks also for the pointers on the UnboudedSource SDF wrapper. > > > Why do you want to evolve a bounded SDF into an unbounded SDF (is the > > restriction truly unbounded)? > > To prove that the idea that we can use the same API/code for both Bounded > and > Unbounded sources holds true, but the more I think about it the harder it > is to > me to get an example that cannot be built with a combination of Unbounded > SDF a > la Watch + a Bounded SDF. Can you think of any? > No but I'm sure someone will think of something eventually. > ... Java has been less of a focus since it is the most mature non-portable > > implementation but hopefully will move in that direction quickly soon. > > We MUST move into this direction if we want to sell the idea of removing > non > portable translations to runner authors. There are few numbers on > performance of > pipelines using portability + the Java SDK Harness but if we take the > ValidatesRunner tests as the only single case where we have numbers > (notice that > VR uses the EMBEDDED environment, the lowest overhead one for Java runs) > the > numbers are far from good from looking at the latest execution times in > Jenkins: > > - Flink Runner VR #6900: 6 mins 22s vs Portable VR #4332 + #4343: 32 mins > (5 times slower) > > - Spark Runner VR #6854: 8 mins 45s vs Portable VR #2322: 27 mins > (3 times slower) > > Of course we can argue that this use case of short lived pipelines with > not big > data is not the average Beam use case, but still the numbers don’t look > good to > sell. I agree. > On Wed, Mar 4, 2020 at 10:52 PM Luke Cwik <lc...@google.com> wrote: > > > > > > > > On Wed, Mar 4, 2020 at 7:36 AM Ismaël Mejía <ieme...@gmail.com> wrote: > >> > >> > I think we should move to a world where *all* runners become portable > runners. > >> > The at doesn't mean they all need to user docker images, or even > GRPC, but I > >> > don't think having classical-only or classical-excluded features is > where we > >> > want to be long-term. > >> > >> Robert I agree 100% with you, I dream of the day where classic runners > do not > >> exist anymore and we do not have this issues like this one (of not > available > >> features), however there are still two requirements to abandon them: > (1) that > >> the performance overhead is not considerable bigger for existing users > (in > >> particular Java users) and (2) that the portability abstractions are > mature. We > >> are getting there, but not yet there. > > > > > > Dataflow and its internal counter point (Flume) have had good experience > running Go and Python portable pipelines at the same or better performance > then the closest non-portable equivalent has been. Java has been less of a > focus since it is the most mature non-portable implementation but hopefully > will move in that direction quickly soon. This would be a great time for > any contributors who are interested in specific runners to help migrate > them to portable implementations. > > > >> > >> On Tue, Mar 3, 2020 at 6:57 PM Robert Bradshaw <rober...@google.com> > wrote: > >> > > >> > On Tue, Mar 3, 2020 at 9:11 AM Ismaël Mejía <ieme...@gmail.com> > wrote: > >> > > > >> > > > the unification of bounded/unbounded within SplittableDoFn has > always been a goal. > >> > > > >> > > I am glad to know that my intuition is correct and that this was > envisioned, the > >> > > idea of checkpoints for bounded inputs sounds super really useful. > Eager to try > >> > > that on practice. > >> > > > >> > > An explicit example (with a WatermarkEstimator for a bounded case > would be > >> > > really nice to see, for learning purposes), also with the > unification goal what > >> > > if we align then the Bounded SDFs to have similar signatures no? I > mean the > >> > > method that returns a continuation even for the Bounded case. > >> > > > >> > > > Currently the watermark that is reported as part of the > PollResult is passed > >> > > > to the ProcessContext.updateWatermark [1, 2] function and instead > that call > >> > > > would be redirected to the ManualWatermarkEstimator.setWatermark > function [3]. > >> > > > >> > > Is there a JIRA for the Watch adjustments so we don't forget to > integrate the > >> > > WatermarkEstimators in? I am really curious on the implementation > to see if I > >> > > finally understand the internals of Watch too. > >> > > > >> > > Extra question: Do you think we can have a naive version of > Unbounded SDF like > >> > > we have the naive one on classical runners (if I understood > correctly the > >> > > current one is only for portable runners). I worry about the > adoption potential. > >> > > >> > I think we should move to a world where *all* runners become portable > >> > runners. The at doesn't mean they all need to user docker images, or > >> > even GRPC, but I don't think having classical-only or > >> > classical-excluded features is where we want to be long-term. > >> > > >> > > On Tue, Mar 3, 2020 at 1:41 AM Robert Bradshaw <rober...@google.com> > wrote: > >> > > > > >> > > > I don't have a strong preference for using a provider/having a > set of > >> > > > tightly coupled methods in Java, other than that we be consistent > (and > >> > > > we already use the methods style for restrictions). > >> > > > > >> > > > On Mon, Mar 2, 2020 at 3:32 PM Luke Cwik <lc...@google.com> > wrote: > >> > > > > > >> > > > > Jan, there are some parts of Apache Beam the watermarks package > will likely rely on (@Experimental annotation, javadoc links) but > fundamentally should not rely on core and someone could create a separate > package for this. > >> > > > > >> > > > I think it does make sense for a set of common watermark trackers > to > >> > > > be shipped with core (e.g. manual, monotonic, and eventually a > >> > > > probabilistic one). > >> > > > > >> > > > > Ismael, the unification of bounded/unbounded within > SplittableDoFn has always been a goal. There are a set of features that > BoundedSources are unlikely to use but would still be allowed to use them. > For example, bounded sources may want to have support for checkpointing > since I could foresee an BoundedSource that can notice that a certain > resource becomes unavailable and can only process it later. The choice of > which watermark estimator to use is a likely point of difference between > bounded and unbounded SDFs since bounded SDFs would be able to use a very > simple estimator where the watermark is held at -infinity and only advances > to +infinity once there is no more data to process. But even though > unbounded SDFs are likely to be the most common users of varied watermark > estimators, a bounded SDF may still want to advance the watermark as they > read records so that runners that are more "streaming" (for example micro > batch) could process the entire pipeline in parallel vs other runners that > execute one whole segment of the pipeline at a time. > >> > > > > >> > > > Put another way, the value of watermark trackers is to allow > >> > > > processing to continue downstream before the source has completed > >> > > > reading. This is of course essential for streaming, but If the > source > >> > > > is read to completion before downstream stages start (as is the > case > >> > > > for most batch runners) it is not needed. What this unification > does > >> > > > allow, however, is a source to be written in such a way that can > be > >> > > > efficiently used in both batch and streaming mode. > >> > > > > >> > > > > Currently the watermark that is reported as part of the > PollResult is passed to the ProcessContext.updateWatermark [1, 2] function > and instead that call would be redirected to the > ManualWatermarkEstimator.setWatermark function [3]. > >> > > > > > >> > > > > 1: > https://github.com/apache/beam/blob/a16725593b84b84b37bc67cd202d1ac8b724c6f4/sdks/java/core/src/main/java/org/apache/beam/sdk/transforms/Watch.java#L757 > >> > > > > 2: > https://github.com/apache/beam/blob/a16725593b84b84b37bc67cd202d1ac8b724c6f4/sdks/java/core/src/main/java/org/apache/beam/sdk/transforms/DoFn.java#L275 > >> > > > > 3: > https://github.com/apache/beam/blob/fb42666a4e1aec0413f161c742d8f010ef9fe9f2/sdks/java/core/src/main/java/org/apache/beam/sdk/transforms/splittabledofn/ManualWatermarkEstimator.java#L45 > >> > > > > > >> > > > > On Fri, Feb 28, 2020 at 6:09 AM Ismaël Mejía <ieme...@gmail.com> > wrote: > >> > > > >> > >> > > > >> I just realized that the HBaseIO example is not a good one > because we can > >> > > > >> already have Watch like behavior as we do for Partition > discovery in HCatalogIO. > >> > > > >> Still I am interested on your views on bounded/unbounded > unification. > >> > > > >> > >> > > > >> Interesting question2: How this will annotations connect with > the Watch > >> > > > >> transform Polling patterns? > >> > > > >> > https://github.com/apache/beam/blob/650e6cd9c707472c34055382a1356cf22d14ee5e/sdks/java/core/src/main/java/org/apache/beam/sdk/transforms/Watch.java#L178 > >> > > > >> > >> > > > >> > >> > > > >> On Fri, Feb 28, 2020 at 10:47 AM Ismaël Mejía < > ieme...@gmail.com> wrote: > >> > > > >>> > >> > > > >>> Really interesting! Implementing correctly the watermark has > been a common > >> > > > >>> struggle for IO authors, to the point that some IOs still > have issues around > >> > > > >>> that. So +1 for this, in particular if we can get to reuse > common patterns. > >> > > > >>> I was not aware of Boyuan's work around this, really nice. > >> > > > >>> > >> > > > >>> One aspect I have always being confused about since I read > the SDF proposal > >> > > > >>> documents is if we could get to have a single API for both > Bounded and Unbounded > >> > > > >>> IO by somehow assuming that with a BoundedSDF is an > UnboundedSDF special case. > >> > > > >>> Could WatermarkEstimator help in this direction? > >> > > > >>> > >> > > > >>> One quick case that I can think is to make the current > HBaseIO SDF to work in an > >> > > > >>> unbounded manner, for example to 'watch and read new tables'. > >> > > > >>> > >> > > > >>> > >> > > > >>> On Thu, Feb 27, 2020 at 11:43 PM Luke Cwik <lc...@google.com> > wrote: > >> > > > >>>> > >> > > > >>>> See this doc[1] and blog[2] for some context about > SplittableDoFns. > >> > > > >>>> > >> > > > >>>> To support watermark reporting within the Java SDK for > SplittableDoFns, we need a way to have SDF authors to report watermark > estimates over the element and restriction pair that they are processing. > >> > > > >>>> > >> > > > >>>> For UnboundedSources, it was found to be a pain point to ask > each SDF author to write their own watermark estimation which typically > prevented re-use. Therefore we would like to have a "library" of watermark > estimators that help SDF authors perform this estimation similar to how > there is a "library" of restrictions and restriction trackers that SDF > authors can use. For SDF authors where the existing library doesn't work, > they can add additional ones that observe timestamps of elements or choose > to directly report the watermark through a "ManualWatermarkEstimator" > parameter that can be supplied to @ProcessElement methods. > >> > > > >>>> > >> > > > >>>> The public facing portion of the DoFn changes adds three new > annotations for new DoFn style methods: > >> > > > >>>> GetInitialWatermarkEstimatorState: Returns the initial > watermark state, similar to GetInitialRestriction > >> > > > >>>> GetWatermarkEstimatorStateCoder: Returns a coder compatible > with watermark state type, similar to GetRestrictionCoder for restrictions > returned by GetInitialRestriction. > >> > > > >>>> NewWatermarkEstimator: Returns a watermark estimator that > either the framework invokes allowing it to observe the timestamps of > output records or a manual watermark estimator that can be explicitly > invoked to update the watermark. > >> > > > >>>> > >> > > > >>>> See [3] for an initial PR with the public facing additions > to the core Java API related to SplittableDoFn. > >> > > > >>>> > >> > > > >>>> This mirrors a bunch of work that was done by Boyuan within > the Pyhon SDK [4, 5] but in the style of new DoFn parameter/method > invocation we have in the Java SDK. > >> > > > >>>> > >> > > > >>>> 1: https://s.apache.org/splittable-do-fn > >> > > > >>>> 2: > https://beam.apache.org/blog/2017/08/16/splittable-do-fn.html > >> > > > >>>> 3: https://github.com/apache/beam/pull/10992 > >> > > > >>>> 4: https://github.com/apache/beam/pull/9794 > >> > > > >>>> 5: https://github.com/apache/beam/pull/10375 >
