Greetings,
When playing around with the following simple event-time stream aggregation:
SELECT
TUMBLE_START(event_time, INTERVAL '1' DAY) as event_time,
...
FROM input
GROUP BY TUMBLE(event_time, INTERVAL '1' DAY), symbol
...to my surprise I found out that the t
Hi there,
The new feature in Flink 1.11 will provide us the Unaligned Checkpoint which
means a operator subtask does not need to wait all the Checkpoint barrier and
will not block some channels. As the Checkpoint barrier is the key mechanism
for Exactly Once guarantee, I am not sure Unaligned C
Hi guys,
I have a few questions on state backends.
Is there a guideline on how big the state has to be where it makes sense to
use RocksDB rather than FsStatebackend ? Is there an analysis on
latency for a full checkpoint for FsSateBackend based on increase in state
size ?
Best,
Nick.
Hi folks,
I want to split my stream for some invalid message handling, and need help
understanding a few things.
Question 1: Why is *split *operator deprecated?
Question 2: Why side-outputs are only supported for ProcessFunction,
KeyedProcessFunction etc.
The doc on side-outputs says: "*You can us
Hi
According to my observation in the hbase community, there are still lots of
hbase users running their production cluster with version 1.x (1.4x or
1.5.x). so I'd like to suggest that
supporting both hbase1.x & hbase2.x connector.
Thanks.
On Sat, Jun 20, 2020 at 2:41 PM Ming Li wrote:
> +1 t
Hi Sergii,
The current watermark strategy is correct.
The window's output is drived by watermark. Before when the window is
triggered, the watermark which triggers
it will be emitted after the result of the window has been fully emitted.
Hence, the watermark won't outpace the
right margin of the
Hi Sergii,
Window operator won't affect/adjust the output watermark, it just
propagated as-is which is said in the document.
I think the mistake here is you are using the wrong event-time of the
window, actually, you should use TUMBLE_ROWTIME(...) as event_time [1].
The event-time of the window sh
Hi Weizheng,
The unaligned checkpoint (UC) only supports exactly-once mode in Flink 1.11
except savepoint mode. The savepoint is probably used in job rescaling
scenario and we plan to support it in future release version. Of course UC can
satisfy exactly-once semantic as promised.
Regarding th
Hi Dhurandar,
With my understand I think what you need is to get notified when a file is
written successfully (committed) on the S3 FileSystem. However, currently there
is no public API for the listener and there an issue tracking it [1].
With the current version, one possible method co
Thank you Zhijiang! The second question about config is just because I find a
method in InputProcessorUtil. I guess AT_LEAST_ONCE mode is a simpler way to
handle checkpont barrier?
private static CheckpointBarrierHandler createCheckpointBarrierHandler(
StreamConfig config,
InputGat
From implementation or logic complication perspective, the AT_LEAST_ONCE is
somehow simpler compared with EXACTLY_ONCE w/o unaligned, since
it can always process data without blocking any channels.
--
From:Lu Weizheng
Send Time:2
您好，jack：
Table API 不用 if/else 直接用类似逻辑即可：
val t1 = table.filter('x > 2).groupBy(..)
val t2 = table.filter('x <= 2).groupBy(..)
t1.insert_into("sink1)
t2.insert_into("sink2")
Best,
Jincheng
jack 于2020年6月19日周五 上午10:35写道：
>
> 测试使用如下结构：
> table= t_env.from_path("source")
>
> if table.filter("
12 matches
Mail list logo
