Hi Seth,

that's not possible with the current interface.
There have been some discussions about how to address issues of idle
sources (or partitions).
Aljoscha (in CC) should know more about that.

Best, Fabian

2017-12-13 18:13 GMT+01:00 Seth Wiesman <swies...@mediamath.com>:

> Quick follow up question. Is there some way to notify a TimestampAssigner
> that is consuming from an idle source?
>
>
>
> <http://www.mediamath.com/>
>
> *Seth Wiesman *| Software Engineer, Data
>
> 4 World Trade Center, 46th Floor, New York, NY 10007
>
>
>
>
>
> *From: *Seth Wiesman <swies...@mediamath.com>
> *Date: *Wednesday, December 13, 2017 at 12:04 PM
> *To: *Timo Walther <twal...@apache.org>, "user@flink.apache.org" <
> user@flink.apache.org>
> *Subject: *Re: Watermark in broadcast
>
>
>
> Hi Timo,
>
>
>
> I think you are correct. This stream is consumed from Kafka and the number
> of partitions is much less than the parallelism of the program so there
> would be many partitions that never forward watermarks greater than
> Long.Min_Value.
>
>
>
> Thank you for the quick response.
>
>
>
> <http://www.mediamath.com/>
>
> *Seth Wiesman *| Software Engineer, Data
>
> 4 World Trade Center, 46th Floor, New York, NY 10007
>
>
>
>
>
> *From: *Timo Walther <twal...@apache.org>
> *Date: *Wednesday, December 13, 2017 at 11:46 AM
> *To: *"user@flink.apache.org" <user@flink.apache.org>
> *Subject: *Re: Watermark in broadcast
>
>
>
> Hi Seth,
>
> are you sure that all partitions of the broadcasted stream send a
> watermark? processWatermark is only called if a minimum watermark arrived
> from all partitions.
>
> Regards,
> Timo
>
> Am 12/13/17 um 5:10 PM schrieb Seth Wiesman:
>
> Hi,
>
>
>
> How are watermarks propagated during a broadcast partition? I have a
> TwoInputStreamTransformation that takes a broadcast stream as one of its
> inputs. Both streams are assigned timestamps and watermarks before being
> connected however I only ever see watermarks from my non-broadcast stream.
> Is this expected behavior? Currently I have overridden processWatermark1 to
> unconditionally call processWatermark but that does not seem like an ideal
> solution.
>
>
>
> Thank you,
>
> <http://www.mediamath.com/>
>
> *Seth Wiesman *| Software Engineer, Data
>
> 4 World Trade Center, 46th Floor, New York, NY 10007
>
>
>
>
>

Reply via email to