This might be a bug in the Flink runner, because it is implemented here
<https://github.com/apache/beam/blob/b87f1f01197a78d60fb08a0ee7a4ef96bc4e08eb/runners/core-java/src/main/java/org/apache/beam/runners/core/InMemoryTimerInternals.java#L164>
.

On Tue, Jul 26, 2022 at 9:14 AM Cristian Constantinescu <[email protected]>
wrote:

> Hi everyone,
>
> Quick question about GroupIntoBatches.
>
> When running on Flink, eventually it hits an unsupported
> exception "Canceling a timer by ID is not yet supported." on this line [1].
> The source inputs are AVRO files for testing (batch) but will use kafka
> topics (streaming) when deployed.
>
> This happens when the batch is filled (10 items) and the max buffering
> time timer needs to be cancelled.
>
> Anyone else observed this issue?
>
> On a related note. Looks like the FlinkStatefulDoFnFunction [2] uses
> InMemoryTimerInternals. Curious why is FlinkTimerInternals not used? I
> would guess there's a difference between batch and streaming requirements?
>
> Thank you,
> Cristian
>
>
> [1]
> https://github.com/apache/beam/blob/b87f1f01197a78d60fb08a0ee7a4ef96bc4e08eb/runners/core-java/src/main/java/org/apache/beam/runners/core/InMemoryTimerInternals.java#L157
> [2]
> https://github.com/apache/beam/blob/b87f1f01197a78d60fb08a0ee7a4ef96bc4e08eb/runners/flink/src/main/java/org/apache/beam/runners/flink/translation/functions/FlinkStatefulDoFnFunction.java#L136
> [3]
> https://github.com/apache/beam/blob/b87f1f01197a78d60fb08a0ee7a4ef96bc4e08eb/runners/flink/src/main/java/org/apache/beam/runners/flink/translation/wrappers/streaming/DoFnOperator.java#L1314
>

Reply via email to