If you have picked up all the changes for SPARK-18057, the Kafka “broker”
supporting v1.0+ should be compatible with Spark's Kafka adapter.

Can you post more details about the “failed to send SSL close message”
errors ?

(The default Kafka version is 2.0.0 in Spark Kafka adapter after SPARK-18057
)

Thanks

On Fri, Aug 17, 2018 at 3:53 PM Basil Hariri
<basil.har...@microsoft.com.invalid> wrote:

> Hi all,
>
>
>
> I work on Azure Event Hubs (Microsoft’s PaaS offering similar to Apache
> Kafka) and am trying to get our new Kafka head
> <https://azure.microsoft.com/en-us/blog/azure-event-hubs-for-kafka-ecosystems-in-public-preview/>
> to play nice with Spark’s Kafka adapter. The goal is for our Kafka endpoint
> to be completely compatible with Spark’s Kafka adapter, but I’m running
> into some issues that I think are related to versioning. I’ve been trying
> to tinker with the kafka-0-10-sql
> <https://github.com/apache/spark/tree/master/external/kafka-0-10-sql> and
> kafka-0-10
> <https://github.com/apache/spark/tree/master/external/kafka-0-10-sql>
> adapters on Github and was wondering if someone could take a second to
> point me in the right direction with:
>
>
>
>    1. What is the difference between those two adapters? My hunch is that
>    kafka-0-10-sql supports structured streaming while kafka-10-0 still uses
>    Spark streaming, but I haven’t found anything to verify that.
>    2. Event Hubs’ Kafka endpoint only supports Kafka 1.0 and later, and
>    the errors I get when trying to connect to Spark (“failed to send SSL close
>    message” / broken pipe errors) have usually shown up when using Kafka v0.10
>    applications with our endpoint. I built from source after I saw that both
>    libraries were updated for Kafka 2.0 support (late last week), but I’m
>    still running into the same issues. Do Spark’s Kafka adapters generally
>    downgrade to Kafka v0.10 protocols? If not, is there any other reason to
>    believe that a Kafka “broker” that doesn’t support v0.10 protocols but
>    supports v1.0+ would be incompatible with Spark’s Kafka adapter?
>
>
>
> Thanks in advance, please let me know if there’s a different place I
> should be posting this
>
>
>
> Sincerely,
>
> Basil
>
>
>

Reply via email to