apex-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jaspal Singh <jaspal.singh1...@gmail.com>
Subject Re: Datatorrent fault tolerance
Date Thu, 06 Oct 2016 22:37:53 GMT
Hi Thomas,

In our case we are writing the results back to maprstreams topic based on
some validations.


Thanks
Jaspal

On Thursday, October 6, 2016, Thomas Weise <thw@apache.org> wrote:

> Hi,
>
> which operators in your application are writing to external systems?
>
> When you look at the example from the blog (https://github.com/
> DataTorrent/examples/tree/master/tutorials/exactly-once), there is Kafka
> input, which is configured to be idempotent. The results are written to
> JDBC. That operator by itself supports exactly-once through transactions
> (in conjunction with idempotent input), hence there is no need to configure
> the processing mode at all.
>
> Thomas
>
>
> On Thu, Oct 6, 2016 at 1:52 PM, Bandaru, Srinivas <
> srinivas.bandaru@optum.com
> <javascript:_e(%7B%7D,'cvml','srinivas.bandaru@optum.com');>> wrote:
>
>> Thanks Thomas. We have set the set all three operator properties, Kafka
>> input operator has set to “Exactly_once” and downstream two operators set
>> as AT_MOST_ONCE same as mentioned in documentation.  We have also referred
>> the documentation you have shared but couldn’t find the right properties
>> for exactly_once scenario. Is there anything we are missing or need to add
>> additional properties if someone else used this functionality.
>>
>>
>>
>> *Any help would be greatly appreciated??*
>>
>>
>>
>> We tried with different combinations and trying to set all three
>> properties with “EXACTLY_ONCE”, But application throwing an with error
>> saying *“If the processing mode for an operator is specified as
>> EXACTLY_ONCE then the processing mode for all downstream operators should
>> be specified as AT_MOST_ONCE otherwise it will result in an error.*”
>> Application behavior is same as the Datatorrent documentation but it’s not
>> working.
>>
>>
>>
>> Thanks,
>>
>> Srinivas
>>
>>
>>
>>
>>
>> *From:* Thomas Weise [mailto:thw@apache.org
>> <javascript:_e(%7B%7D,'cvml','thw@apache.org');>]
>> *Sent:* Thursday, October 06, 2016 3:03 PM
>> *To:* users@apex.apache.org
>> <javascript:_e(%7B%7D,'cvml','users@apex.apache.org');>
>> *Subject:* Re: Datatorrent fault tolerance
>>
>>
>>
>> Hi,
>>
>>
>>
>> It would be necessary to know a bit more about your application for
>> specific recommendations, but from what I see above, a few things don't
>> look right.
>>
>>
>>
>> It appears that you are setting the processing mode on the input
>> operator, which only reads from Kafka. Exactly-once is important for
>> operators that affect the state of external systems, for example when
>> writing to a database, or, producing messages to Kafka.
>>
>>
>>
>> There is also a problem with the documentation that may contribute to the
>> confusion. The processing mode you are trying to use isn't really
>> exactly-once and probably will be deprecated in a future release. Here are
>> some better resources on fault tolerance and processing semantics:
>>
>>
>>
>> http://apex.apache.org/docs.html
>>
>> https://www.youtube.com/watch?v=FCMY6Ii89Nw
>>
>> http://www.slideshare.net/ApacheApexOrganizer/webinar-fault-
>> toleranceandprocessingsemantics
>>
>> https://www.datatorrent.com/blog/end-to-end-exactly-once-wit
>> h-apache-apex/
>>
>>
>>
>> HTH,
>>
>> Thomas
>>
>>
>>
>>
>>
>> On Thu, Oct 6, 2016 at 11:26 AM, Bandaru, Srinivas <
>> srinivas.bandaru@optum.com
>> <javascript:_e(%7B%7D,'cvml','srinivas.bandaru@optum.com');>> wrote:
>>
>> Hi, Need some help!
>>
>> Need your advice in implementing the fault tolerance mechanism using
>> datatorrent. Currently we are trying to implement with EXACTLY_ONCE
>> scenario.
>>
>> We are referring the below documentation. The highlighted line says if we
>> need to use the EXACTLY_ONCE processing mode, the downstream operators
>> should have AT_MOST_ONCE. We added the operator level attributes in the
>> properties.xml (attached below) and when we launch the application, no
>> messages are coming through the last operator *topicUpdate *but we are
>> able to see the messages when we comment those attributes.
>>
>> *Please have a look at it, either we are specifying the attributes
>> incorrectly or is there any additional step to be done*.
>>
>>
>>
>>
>>
>> ·         *PROCESSING_MODE*
>>
>> static final Attribute <https://www.datatorrent.com/docs/apidocs/com/datatorrent/api/Attribute.html><Operator.ProcessingMode
<https://www.datatorrent.com/docs/apidocs/com/datatorrent/api/Operator.ProcessingMode.html>>
PROCESSING_MODE
>>
>> The payload processing mode for this operator - at most once, exactly
>> once, or default at least once. If the processing mode for an operator is
>> specified as AT_MOST_ONCE and no processing mode is specified for the
>> downstream operators if any, the processing mode of the downstream
>> operators is automatically set to AT_MOST_ONCE. If a different processing
>> mode is specified for the downstream operators it will result in an error. If
>> the processing mode for an operator is specified as EXACTLY_ONCE then the
>> processing mode for all downstream operators should be specified as
>> AT_MOST_ONCE otherwise it will result in an error.
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> Thanks,
>>
>> Srinivas
>>
>>
>> This e-mail, including attachments, may include confidential and/or
>> proprietary information, and may be used only by the person or entity
>> to which it is addressed. If the reader of this e-mail is not the intended
>> recipient or his or her authorized agent, the reader is hereby notified
>> that any dissemination, distribution or copying of this e-mail is
>> prohibited. If you have received this e-mail in error, please notify the
>> sender by replying to this message and delete this e-mail immediately.
>>
>>
>>
>>
>> This e-mail, including attachments, may include confidential and/or
>> proprietary information, and may be used only by the person or entity
>> to which it is addressed. If the reader of this e-mail is not the intended
>> recipient or his or her authorized agent, the reader is hereby notified
>> that any dissemination, distribution or copying of this e-mail is
>> prohibited. If you have received this e-mail in error, please notify the
>> sender by replying to this message and delete this e-mail immediately.
>>
>
>

Mime
View raw message