Thanks for the update, Mang and Ron,
The new proposal looks good to me in general, especially keeping the
behavior
consistent between batch and streaming mode by default. This is how we do
it
in the previous "table.dml-sync" option on ML [1].
Besides that, I just have some final minor comments re
退订
Hi Piotr,
Thanks for the further questions.
To answer both your questions, let's consider the different flows for
interaction between RateLimitingStrategy and AsyncSinkWriter (IMO it is
important since it affects the interface!):
One potential interaction flow is:
1. AsyncSinkWriter construct
jinshuangxian created FLINK-28251:
-
Summary: flink-avro format reader may thrown
’java.util.NoSuchElementException' Exception in some cases
Key: FLINK-28251
URL: https://issues.apache.org/jira/browse/FLINK-282
jinshuangxian created FLINK-28250:
-
Summary: exactly-once sink kafka cause out of memory
Key: FLINK-28250
URL: https://issues.apache.org/jira/browse/FLINK-28250
Project: Flink
Issue Type: Bug