Hi Chesnay,
Thank you for reply.
I figure out that issue with using livenessProbe on Task Manager
deployment. But I think it is still a workaround.
I am using Flink 1.9.1 (currently its latest version)
And I am getting "connection unexpectedly closed by remote task manager"
error on Task Manager.
Hi Shuwen,
> When to call close() ? After every element processed? Or on
> ProcessFunction.close() ? Or never to use it?
IMO, the #close() function is used to manage the lifecycle of #Collector
instead of a single element. I think it should not be called in user function
unless you have
Hi Community,
In ProcessFunction class, ProcessElement function, there is a Collector
that has 2 method: collect() and close(). I would like to know:
1. When to call close() ? After every element processed? Or
on ProcessFunction.close() ? Or never to use it? If it's been closed
already, can the
Hi Akash,
Flink doesn't support auto scaling in core currently, it may be supported
in the future, when the new scheduling architecture is implemented
https://issues.apache.org/jira/browse/FLINK-10407 .
You can do it externally by cancel the job with a savepoint, update the
parallelism, and
Hi Akash,
You can use Pravega connector to integrate with Flink, the source code is
here[1].
In short, relying on its rescalable state feature[2] flink supports
scalable streaming jobs.
Currently, the mainstream solution about auto-scaling is Flink + K8S, I can
share some resources with you[3].
Hi,
Does Flunk support auto scaling. I read that it is supported using pravega?
Is it incorporated in any version.
Thanks,
Akash Goel
Just to add up, if you use LocalStreamEnvironment, you can pass a
configuration and you can set "execution.savepoint.path" to point to your
savepoint.
Best,
Arvid
On Wed, Nov 27, 2019 at 1:00 PM Congxian Qiu wrote:
> Hi,
>
> You can recovery from checkpoint/savepoint if JM can read from the
>
Could you try accessing :/#/overview ?
The REST API is obviously accessible, and hence the WebUI should be too.
How did you setup the session cluster? Are you using some custom Flink
build or something, which potentially excluded flink-runtime-web from
the classpath?
On 28/11/2019 10:02,
Thank you all for investigation/reporting/discussion. I have merged an older PR
[1] that was fixing this issue which was previously rejected as we didn’t
realise this is a production issue.
I have merged it and issue should be fixed in Flink 1.10, 1.9.2 and 1.8.3
releases.
Piotrek
[1]
The akka.watch configuration options haven't been used for a while
irrespective of FLINK-13883 (but I can't quite tell atm since when).
Let's start with what version of Flink you are using, and what the
taskmanager/jobmanager logs say.
On 25/11/2019 12:05, Eray Arslan wrote:
Hi,
I have
Hi Harrison,
One thing to keep in mind is that Flink will only write files if there
is data to write. If, for example, your partition is not active for a
period of time, then no files will be written.
Could this explain the fact that dt 2019-11-24T01 and 2019-11-24T02
are entirely skipped?
In
Hi Li,
S3 file sink will write data into prefixes, with as many part-files as the
degree of parallelism. This structure comes from the good ol' Hadoop days,
where an output folder also contained part-files and is independent of S3.
However, each of the part-files will be uploaded in a multipart
What do you mean by "from within a sink"? Do you have a custom sink?
If you want to write to different Kafka topics from the same sink, you can
do that using a custom KafkaSerializationSchema. It allows you to return a
ProducerRecord with a custom target topic set. (A Kafka sink can write to
Hi
Checkpooint 不会导致 State 被清理。Checkpoint 只是将 State 备份到一个远程存储,供后续恢复使用。如果有大量
CopyOnWriteStateTable$StateTableEntry 的话,首先需要确认,你作业真的会有大量的 state 存在吗?另外
Checkpoint 相关的配置是什么,OOM 前是否有 checkpoint 失败这类的
Best,
Congxian
李茂伟 于2019年11月28日周四 下午9:02写道:
> hi all ~
>
>
hi all ~
我在使用FsStateBackend,这种方式checkpoint时,导致oom,查看堆内存中有大量的CopyOnWriteStateTable$StateTableEntry对象没有被清理。请问在使用FsStateBackend这种方式checkpoint后,内存中的state会被清理吗?
Hi Rahul,
can you check if the KuduSink tests of Apache Bahir shed any light? [1]
Best,
Arvid
[1]
https://github.com/apache/bahir-flink/blob/55240a993df999d66aefa36e587be719c29be92a/flink-connector-kudu/src/test/java/org/apache/flink/connectors/kudu/streaming/KuduSinkTest.java
On Mon, Nov
Hi Avi,
it seems to me that you are not really needing any split feature. As far as
I can see in your picture you want to apply two different windows on the
same input data.
In that case you simply use two different subgraphs.
stream = ...
stream1 = stream.window(...).addSink()
stream2 =
the chk-* directory is not found , I think the misssing because of jobmanager
removes it automaticly , but why it still in zookeeper?
----
??:"Vijay Bhaskar"http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/
Hi Victor,
you could implement your own SinkFunction that wraps the KafkaProducer.
However, since you may need to check if the write operation is successful,
you probably need to subclass KafkaProducer and implement your own error
handling.
Best,
Arvid
On Mon, Nov 25, 2019 at 7:51 AM vino yang
Hi Lei,
if you use
public JobExecutionResult StreamExecutionEnvironment#execute()
You can retrieve the job id through the result.
result.getJobID()
Best,
Arvid
On Mon, Nov 25, 2019 at 3:50 AM Ana wrote:
> Hi Lei,
>
> To add, you may use Hadoop Resource Manager REST APIs
>
One more thing:
You configured:
high-availability.cluster-id: /cluster-test
it should be:
high-availability.cluster-id: cluster-test
I don't think this is major issue, in case it helps, you can check.
Can you check one more thing:
Is check pointing happening or not?
Were you able to see the chk-*
hi??
Is there any deference ??for me using nas is more convenient to test
currently???
from the docs seems hdfs ,s3, nfs etc all will be fine.
----
??:"vino yang"http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/
Hi,
Why do you not use HDFS directly?
Best,
Vino
曾祥才 于2019年11月28日周四 下午6:48写道:
>
> anyone have the same problem? pls help, thks
>
>
>
> -- 原始邮件 --
> *发件人:* "曾祥才";
> *发送时间:* 2019年11月28日(星期四) 下午2:46
> *收件人:* "Vijay Bhaskar";
> *抄送:* "User-Flink";
> *主题:* 回复:
Hi Jatin,
Which version do you use?
Best,
Vino
Jatin Banger 于2019年11月28日周四 下午5:03写道:
> Hi,
>
> I checked the log file there is no error.
> And I checked the pods internal ports by using rest api.
>
> # curl : 4081
> {"errors":["Not found."]}
> 4081 is the Ui port
>
> # curl :4081/config
>
anyone have the same problem?? pls help, thks
----
??:"??"http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/
Hi all,我在使用flink cep的过程中遇到了如下问题(一个 A FollowedBy B 的 case)
Case1. 我构造的数据源中全是事件A,全速发送
该任务在执行数秒后阻塞(task之间不再有数据交换),源端的backpressure=1,一段时间后TM失去心跳响应,任务异常终止,此时TaskManager进程存活,但通过jstat观察内存使用情况后,发现内存被耗尽,且不停的FullGC
Case2. 我构造的数据源中交叉发送事件A和事件B,全速发送
大家好:
我在官方文档中看到了/jobs/:jobid/stop这个接口,文档上说这个接口可以生成savepoint。因此我想使用这个接口停止运行在yarn上的flink作业,但是会报错。
请问是不是这个接口当前还用不了?是否有其他代替的接口可用?
接口地址我尝试过:
http://host:8088/proxy/application_1572228642684_1907/v1/jobs/a1f207a50cdebd531824d9a4402162ca/stop
和
Hi,
I checked the log file there is no error.
And I checked the pods internal ports by using rest api.
# curl : 4081
{"errors":["Not found."]}
4081 is the Ui port
# curl :4081/config
{"refresh-interval":3000,"timezone-name":"Coordinated Universal
Hi Yingjie,
I read this post and the next one as well (
https://flink.apache.org/2019/07/23/flink-network-stack-2.html).
I mean the bandwidth of the channels between two physical operators. When
they are in different hosts, so when the channels are a network channel.
Thanks
*--*
*-- Felipe
29 matches
Mail list logo