Hi David, sorry for my late reply. I just found time to look into the problem. You were right with your observation that the CEP operator did not behave as I've described it. The problem was that the time of the underlying NFA was not advanced if there were no events buffered in the CEP operator when a new watermark arrived. This was not intended and I opened a PR [1] to fix this problem. I've tested the fix with your example program and it seems to solve the problem that you don't see timeouts after the timeout interval has passed. Thanks for reporting this problem and please excuse my long response time.
Btw, I'll merge the PR this evening. So it should be included in the current snapshot version by the end of tomorrow. [1] https://github.com/apache/flink/pull/2771 Cheers, Till On Fri, Oct 14, 2016 at 11:40 AM, Till Rohrmann <trohrm...@apache.org> wrote: > Hi guys, > > I'll try to come up with an example illustrating the behaviour over the > weekend. > > Cheers, > Till > > On Fri, Oct 14, 2016 at 11:16 AM, David Koch <ogd...@googlemail.com> > wrote: > >> Hello, >> >> Thanks for the code Sameer. Unfortunately, it didn't solve the issue. >> Compared to what I did the principle is the same - make sure that the >> watermark advances even without events present to trigger timeouts in CEP >> patterns. >> >> If Till or anyone else could provide a minimal example illustrating the >> supposed behaviour of: >> >> [CEP] timeout will be detected when the first watermark exceeding the >>> timeout value is received >> >> >> I'd very much appreciate it. >> >> Regards, >> >> David >> >> >> On Wed, Oct 12, 2016 at 1:54 AM, Sameer W <sam...@axiomine.com> wrote: >> >>> Try this. Your WM's need to move forward. Also don't use System >>> Timestamp. Use the timestamp of the element seen as the reference as the >>> elements are most likely lagging the system timestamp. >>> >>> DataStream<Event> withTimestampsAndWatermarks = tuples >>> .assignTimestampsAndWatermarks(new >>> AssignerWithPeriodicWatermarks<Event>() { >>> >>> long waterMarkTmst; >>> long lastEmittedWM=0; >>> @Override >>> public long extractTimestamp(Event element, long >>> previousElementTimestamp) { >>> if(element.tmst>lastEmittedWM){ >>> waterMarkTmst = element.tmst-1; //Assumes increasing >>> timestamps. Need to subtract 1 as more elements with same TS might arrive >>> } >>> return element.tmst; >>> } >>> >>> @Override >>> public Watermark getCurrentWatermark() { >>> if(lastEmittedWM==waterMarkTmst){ //No new event seen, >>> move the WM forward by auto watermark interval >>> waterMarkTmst = waterMarkTmst + 1000l//Increase by >>> auto watermark interval (Watermarks only move forward in time) >>> } >>> lastEmittedWM = waterMarkTmst >>> >>> System.out.println(String.format("Watermark at %s", new >>> Date(waterMarkTmst))); >>> return new Watermark(waterMarkTmst);//Until an event is >>> seem WM==0 starts advancing by 1000ms until an event is seen >>> } >>> }).keyBy("key"); >>> >>> On Tue, Oct 11, 2016 at 7:29 PM, David Koch <ogd...@googlemail.com> >>> wrote: >>> >>>> Hello, >>>> >>>> I tried setting the watermark to System.currentTimeMillis() - 5000L, >>>> event timestamps are System.currentTimeMillis(). I do not observe the >>>> expected behaviour of the PatternTimeoutFunction firing once the watermark >>>> moves past the timeout "anchored" by a pattern match. >>>> >>>> Here is the complete test class source <http://pastebin.com/9WxGq2wv>, >>>> in case someone is interested. The timestamp/watermark assigner looks like >>>> this: >>>> >>>> DataStream<Event> withTimestampsAndWatermarks = tuples >>>> .assignTimestampsAndWatermarks(new >>>> AssignerWithPeriodicWatermarks<Event>() { >>>> >>>> long waterMarkTmst; >>>> >>>> @Override >>>> public long extractTimestamp(Event element, long >>>> previousElementTimestamp) { >>>> return element.tmst; >>>> } >>>> >>>> @Override >>>> public Watermark getCurrentWatermark() { >>>> waterMarkTmst = System.currentTimeMillis() - 5000L; >>>> System.out.println(String.format("Watermark at %s", >>>> new Date(waterMarkTmst))); >>>> return new Watermark(waterMarkTmst); >>>> } >>>> }).keyBy("key"); >>>> >>>> withTimestampsAndWatermarks.getExecutionConfig().setAutoWate >>>> rmarkInterval(1000L); >>>> >>>> // Apply pattern filtering on stream. >>>> PatternStream<Event> patternStream = >>>> CEP.pattern(withTimestampsAndWatermarks, >>>> pattern); >>>> >>>> Any idea what's wrong? >>>> >>>> David >>>> >>>> >>>> On Tue, Oct 11, 2016 at 10:20 PM, Sameer W <sam...@axiomine.com> wrote: >>>> >>>>> Assuming an element with timestamp which is later than the last >>>>> emitted watermark arrives, would it just be dropped because the >>>>> PatternStream does not have a max allowed lateness method? In that case it >>>>> appears that CEP cannot handle late events yet out of the box. >>>>> >>>>> If we do want to support late events can we chain a >>>>> keyBy().timeWindow().allowedLateness(x).map().assignTimestampsAndWatermarks().keyBy() >>>>> again before handing it to the CEP operator. This way we may have the >>>>> patterns fired multiple times but it allows an event to be late and out of >>>>> order. It looks like it will work but is there a less convoluted way. >>>>> >>>>> Thanks, >>>>> Sameer >>>>> >>>>> On Tue, Oct 11, 2016 at 12:17 PM, Till Rohrmann < >>>>> till.rohrm...@gmail.com> wrote: >>>>> >>>>>> But then no element later than the last emitted watermark must be >>>>>> issued by the sources. If that is the case, then this solution should >>>>>> work. >>>>>> >>>>>> Cheers, >>>>>> Till >>>>>> >>>>>> On Tue, Oct 11, 2016 at 4:50 PM, Sameer W <sam...@axiomine.com> >>>>>> wrote: >>>>>> >>>>>>> Hi, >>>>>>> >>>>>>> If you know that the events are arriving in order and a consistent >>>>>>> lag, why not just increment the watermark time every time the >>>>>>> getCurrentWatermark() method is invoked based on the >>>>>>> autoWatermarkInterval >>>>>>> (or less to be conservative). >>>>>>> >>>>>>> You can check if the watermark has changed since the arrival of the >>>>>>> last event and if not increment it in the getCurrentWatermark() method. >>>>>>> Otherwise the watermark will never increase until an element arrive and >>>>>>> if >>>>>>> the stream partition stalls for some reason the whole pipeline freezes. >>>>>>> >>>>>>> Sameer >>>>>>> >>>>>>> >>>>>>> On Tue, Oct 11, 2016 at 6:04 AM, Till Rohrmann < >>>>>>> till.rohrm...@gmail.com> wrote: >>>>>>> >>>>>>>> Hi David, >>>>>>>> >>>>>>>> the problem is still that there is no corresponding watermark >>>>>>>> saying that 4 seconds have now passed. With your code, watermarks will >>>>>>>> be >>>>>>>> periodically emitted but the same watermark will be emitted until a new >>>>>>>> element arrives which will reset the watermark. Thus, the system can >>>>>>>> never >>>>>>>> know until this watermark is seen whether there will be an earlier >>>>>>>> event or >>>>>>>> not. I fear that this is a fundamental problem with stream processing. >>>>>>>> >>>>>>>> You're right that the negation operator won't solve the problem. It >>>>>>>> will indeed suffer from the same problem. >>>>>>>> >>>>>>>> Cheers, >>>>>>>> Till >>>>>>>> >>>>>>>> On Sun, Oct 9, 2016 at 7:37 PM, <lg...@yahoo.com> wrote: >>>>>>>> >>>>>>>>> >>FLINK-3320 <https://issues.apache.org/jira/browse/FLINK-3320> (CEP >>>>>>>>> "not" operator) does not address this because again, how would the >>>>>>>>> "not >>>>>>>>> match" be triggered if no event at all occurs? >>>>>>>>> >>>>>>>>> Good question. >>>>>>>>> >>>>>>>>> I'm not sure whether the following will work: >>>>>>>>> >>>>>>>>> This could be done by creating a CEP matching pattern that uses >>>>>>>>> both of "notNext" (or "notFollowedBy") and "within" constructs. >>>>>>>>> Something >>>>>>>>> like this: >>>>>>>>> >>>>>>>>> Pattern<Event, ?> pattern = Pattern.<Event>begin("first") >>>>>>>>> .notNext("second") >>>>>>>>> .within(Time.seconds(3)); >>>>>>>>> >>>>>>>>> I'm hoping Flink CEP experts (Till?) will comment on this. >>>>>>>>> >>>>>>>>> Note: I have requested these negation patterns to be implemented >>>>>>>>> in Flink CEP, but notNext/notFollowedBy are not yet implemented in >>>>>>>>> Flink.. >>>>>>>>> >>>>>>>>> >>>>>>>>> - LF >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> ------------------------------ >>>>>>>>> *From:* David Koch <ogd...@googlemail.com> >>>>>>>>> *To:* user@flink.apache.org; lg...@yahoo.com >>>>>>>>> *Sent:* Sunday, October 9, 2016 5:51 AM >>>>>>>>> >>>>>>>>> *Subject:* Re: Listening to timed-out patterns in Flink CEP >>>>>>>>> >>>>>>>>> Hello, >>>>>>>>> >>>>>>>>> Thank you for the explanation as well as the link to the other >>>>>>>>> post. Interesting to learn about some of the open JIRAs. >>>>>>>>> >>>>>>>>> Indeed, I was not using event time, but processing time. However, >>>>>>>>> even when using event time I only get notified of timeouts upon >>>>>>>>> subsequent >>>>>>>>> events. >>>>>>>>> >>>>>>>>> The link <http://pastebin.com/x4m3RHQz> contains an example where >>>>>>>>> I read <key> <value> from a socket, wrap this in a custom "event" with >>>>>>>>> timestamp, key the resultant stream by <key> and attempt to detect >>>>>>>>> <key> >>>>>>>>> instances no further than 3 seconds apart using CEP. >>>>>>>>> >>>>>>>>> Apart from the fact that results are only printed when I close the >>>>>>>>> socket (normal?) I don't observe any change in behaviour >>>>>>>>> >>>>>>>>> So event-time/watermarks or not: SOME event has to occur for the >>>>>>>>> timeout to be triggered. >>>>>>>>> >>>>>>>>> FLINK-3320 <https://issues.apache.org/jira/browse/FLINK-3320> (CEP >>>>>>>>> "not" operator) does not address this because again, how would the >>>>>>>>> "not >>>>>>>>> match" be triggered if no event at all occurs? >>>>>>>>> >>>>>>>>> On Sat, Oct 8, 2016 at 12:50 AM, <lg...@yahoo.com> wrote: >>>>>>>>> >>>>>>>>> The following is a better link: >>>>>>>>> >>>>>>>>> http://mail-archives.apache. org/mod_mbox/flink-user/ >>>>>>>>> 201609.mbox/%3CCAC27z% 3DOTtv7USYUm82bE43- >>>>>>>>> DkoGfVC4UAWD6uQwwRgTsE5be8g% >>>>>>>>> 40mail.gmail.com%3E >>>>>>>>> <http://mail-archives.apache.org/mod_mbox/flink-user/201609.mbox/%3CCAC27z%3DOTtv7USYUm82bE43-DkoGfVC4UAWD6uQwwRgTsE5be8g%40mail.gmail.com%3E> >>>>>>>>> >>>>>>>>> >>>>>>>>> - LF >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> ------------------------------ >>>>>>>>> *From:* "lg...@yahoo.com" <lg...@yahoo.com> >>>>>>>>> *To:* "user@flink.apache.org" <user@flink.apache.org> >>>>>>>>> *Sent:* Friday, October 7, 2016 3:36 PM >>>>>>>>> >>>>>>>>> *Subject:* Re: Listening to timed-out patterns in Flink CEP >>>>>>>>> >>>>>>>>> Isn't the upcoming CEP negation (absence of an event) feature >>>>>>>>> solve this issue? >>>>>>>>> >>>>>>>>> See this discussion thread: >>>>>>>>> http://mail-archives.apache. org/mod_mbox/flink-user/ >>>>>>>>> 201609.mbox/%3CCAC27z%3DOD% 2BTq8twBw_ 1YKni5sWAU3g1S9WDpJw0DUwgiG9YX >>>>>>>>> 9Fg%40mail.gmail.com%3E >>>>>>>>> <http://mail-archives.apache.org/mod_mbox/flink-user/201609.mbox/%3CCAC27z%3DOD%2BTq8twBw_1YKni5sWAU3g1S9WDpJw0DUwgiG9YX9Fg%40mail.gmail.com%3E> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> // Atul >>>>>>>>> >>>>>>>>> >>>>>>>>> ------------------------------ >>>>>>>>> *From:* Till Rohrmann <trohrm...@apache.org> >>>>>>>>> *To:* user@flink.apache.org >>>>>>>>> *Sent:* Friday, October 7, 2016 12:58 AM >>>>>>>>> *Subject:* Re: Listening to timed-out patterns in Flink CEP >>>>>>>>> >>>>>>>>> Hi David, >>>>>>>>> >>>>>>>>> in case of event time, the timeout will be detected when the first >>>>>>>>> watermark exceeding the timeout value is received. Thus, it depends a >>>>>>>>> little bit how you generate watermarks (e.g. periodically, watermark >>>>>>>>> per >>>>>>>>> event). >>>>>>>>> >>>>>>>>> In case of processing time, the time is only updated whenever a >>>>>>>>> new element arrives. Thus, if you have an element arriving 4 seconds >>>>>>>>> after >>>>>>>>> Event A, it should detect the timeout. If the next event arrives 20 >>>>>>>>> seconds >>>>>>>>> later, than you won't see the timeout until then. >>>>>>>>> >>>>>>>>> In the case of processing time, we could think about registering >>>>>>>>> timeout timers for processing time. However, I would highly recommend >>>>>>>>> you >>>>>>>>> to use event time, because with processing time, Flink cannot >>>>>>>>> guarantee >>>>>>>>> meaningful computations, because the events might arrive out of order. >>>>>>>>> >>>>>>>>> Cheers, >>>>>>>>> Till >>>>>>>>> >>>>>>>>> On Thu, Oct 6, 2016 at 3:08 PM, David Koch <ogd...@googlemail.com> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> Hello, >>>>>>>>> >>>>>>>>> With Flink CEP, is there a way to actively listen to pattern >>>>>>>>> matches that time out? I am under the impression that this is not >>>>>>>>> possible. >>>>>>>>> >>>>>>>>> In my case I partition a stream containing user web navigation by >>>>>>>>> "userId" to look for sequences of Event A, followed by B within 4 >>>>>>>>> seconds >>>>>>>>> for each user. >>>>>>>>> >>>>>>>>> I registered a PatternTimeoutFunction which assuming a non-match >>>>>>>>> only fires upon the first event after the specified timeout. For >>>>>>>>> example, >>>>>>>>> given user X: Event A, 20 seconds later Event B (or any other type of >>>>>>>>> event). >>>>>>>>> >>>>>>>>> I'd rather have a notification fire directly upon the 4 second >>>>>>>>> interval expiring since passive invalidation is not really applicable >>>>>>>>> in my >>>>>>>>> case. >>>>>>>>> >>>>>>>>> How, if at all can this be achieved with Flink CEP? >>>>>>>>> >>>>>>>>> Thanks, >>>>>>>>> >>>>>>>>> David >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>> >>>>>>> >>>>>> >>>>> >>>> >>> >> >
