Hi
>From my experience, you can first check the jobmanager.log, find out
whether the checkpoint expired or was declined by some task, if expired,
you can follow the adivce of seeksst given above(maybe enable debug log can
help you here), if was declined, then you can go to the taskmanager.log to
Hi
As far as I know, the latency-tracking feature is for debugging usages, you
can use it to debug, and disable it when running the job on production.
>From my side, use $current_processing - $event_time is something ok, but
keep the things in mind: the event time may not be the time ingested in
Hi
从报错来看,你用 StateProcessAPI,StateProcessAPI 的某些 function(这里是
getMetricGroup) 不提供支持,因此会有这个提示,如果你没有显示调用这个 function 的话,那可能是个 bug
Best,
Congxian
guanyq 于2020年3月25日周三 下午3:24写道:
>
>
>
>
>
>
>
Hi
这个地方我理解是,每次处理一定数量的 StateEntry 之后,会获取当前的 timestamp 然后在 RocksDB 的 compaction
时对所有的 StateEntry 进行 filter。
> Calling of TTL filter during compaction slows it down.
Best,
Congxian
LakeShen 于2020年3月26日周四 下午8:55写道:
> Hi 社区的小伙伴,
>
> 我现在有个关于 Flink 1.10 RocksDBStateBackend 状态清理机制的问题,在 1.10中,RocksDB
Jark, 这个功能我们用的还挺多的~
现在还有个痛点是window operator不支持retract输入,所以用了emit就没有办法做到窗口级联使用了。
Jark Wu 于2020年3月27日周五 下午8:01写道:
> Benchao 可以啊。这么隐秘的实验性功能都被你发现了 :D
>
> On Fri, 27 Mar 2020 at 15:24, Benchao Li wrote:
>
> > Hi,
> >
> > 对于第二个场景,可以尝试一下fast emit:
> > table.exec.emit.early-fire.enabled = true
> >
Hi
对于问题 1 在反压的情况下,可能导致 Savepoint 做不成功从而超时,这个暂时没法解决,现在有一个 issue[1] 在做 Unalign
Checkpoint 可以解决反压情况下的 checkpoint
对于问题 3,checkpoint 超时了,超时的定义:在设置的时间内(比如你这里 5 分钟),有 task 没有完成
snapshot。调长超时时间能够一定的缓解这个问题,不过你最好找到超时的原因,然后针对性的优化。
[1] https://issues.apache.org/jira/browse/FLINK-14551
Best,
Congxian
第一个场景: 从SQL的角度,增加时间字段精确到分钟为key,格式如-MM-dd HH:mm。这样是不是就可以实现你要到效果了。
flink小猪 <18579099...@163.com> 于2020年3月27日周五 上午11:29写道:
> 我有两个需求
> 1.如果我需要开一个一分钟的窗口,但是我希望能够每5分钟输出一次结果(输出每分钟的聚合结果,即输出5行聚合结果),该怎么办?
> 2.如果我需要开一个一个小时的窗口,但是我希望能够每5分钟输出一次结果(窗口数据还没全部到达,先输出当前已经聚合的结果),该怎么办?
--
Hi,感谢大家的回复,经过我的分析和测试,我猜测是和taskmanager.network.blocking-shuffle.type=mmap
有关。我看了下监控,Mappred占用的内存会逐渐升至20多G甚至更高,从而导致超过yarn的限制被杀掉。另一方面,如果我配置成taskmanager.network.blocking-shuffle.type=file,监控Mappred一直为0,最后报错会是OutOfMemoryError:direct
buffer memory 说明mmap和file用的是不同的存储。
Hi,
I am looking for end to end latency monitoring of link job. Based on my
study, I have two options:
1. flink provide a latency tracking feature. However, the documentation
says it cannot show actual latency of business logic as it will bypass all
operators.
I've debug it a little bit and I found that it fails in
InstantiationUtil.readObjectFromConfig method when we execute
byte[] bytes = config.getBytes(key, (byte[])null); This returns null.
The key that it is looking for is "edgesInOrder". In the config map, there
are only two entries though.
For
Hi,
Im trying to test my RichAsyncFunction implementation with
OneInputStreamOperatorTestHarness based on [1]. I'm using Flink 1.9.2
My test setup is:
this.processFunction = new MyRichAsyncFunction();
this.testHarness = new OneInputStreamOperatorTestHarness<>(
new
Hi Jark,
Many thanks for creating the issue on Jira and nice summarization :-)
Best,
Dongwon
On Sat, Mar 28, 2020 at 12:37 AM Jark Wu wrote:
> Hi Dongwon,
>
> I saw many requirements on this and I'm big +1 for this.
> I created https://issues.apache.org/jira/browse/FLINK-16833 to track this
>
Hi Dongwon,
I saw many requirements on this and I'm big +1 for this.
I created https://issues.apache.org/jira/browse/FLINK-16833 to track this
effort. Hope this can be done before 1.11 release.
Best,
Jark
On Fri, 27 Mar 2020 at 22:22, Dongwon Kim wrote:
> Hi, I tried flink-jdbc [1] to read
Could you also check the jobmanager logs whether the flink akka is also
bound to
and listening at the hostname "prod-bigd-dn11"? Otherwise, all the package
from
taskmanager will be discarded.
Best,
Yang
Vitaliy Semochkin 于2020年3月27日周五 下午3:35写道:
> Hello Zhu,
>
> The host can be resolved and
Hi, I tried flink-jdbc [1] to read data from Druid because Druid implements
Calcite Avatica [2], but the connection string, jdbc:avatica:remote:url=
http://BROKER:8082/druid/v2/sql/avatica/, is not supported by any of
JDBCDialects [3].
I implement custom JDBCDialect [4], custom
Hi KristoffSC,
the short answer is: you have probably differently configured logger. They
log in a different format or level.
The longer answer: all source connectors currently use the legacy source
thread. That will only change with FLIP-27 [1] being widely adapted. It was
originally planned to
Benchao 可以啊。这么隐秘的实验性功能都被你发现了 :D
On Fri, 27 Mar 2020 at 15:24, Benchao Li wrote:
> Hi,
>
> 对于第二个场景,可以尝试一下fast emit:
> table.exec.emit.early-fire.enabled = true
> table.exec.emit.early-fire.delay = 5min
>
> PS:
> 1. 这个feature并没有在官方文档里面写出来,目前应该是一个实验性质的feature
> 2.
Hi,
据我所知现在还不能直接支持Oracle的driver吧?你是怎么使用Flink SQL读写oracle的哈?
在 2020-03-27 17:21:21,"111" 写道:
>Hi,
>在使用Flink SQL读写Oracle JDBC表时,遇到了timestamp转换异常:
>Caused by: java.lang.ClassCastException: oracle.sql.TIMESTAMP cannot be cast
>to java.sql.Timestamp at
Hi all,
When I run Flink from IDE i can see this prefix in logs
"Legacy Source Thread"
Running the same job as JobCluster on docker, this prefix is not present.
What this prefix means?
Btw, I'm using [1] as ActiveMQ connector.
Thanks.
[1]
Hi, there.
In release-1.10, the memory setup of task managers has changed a lot.
I would like to provide here a third-party tool to simulate and get
the calculation result of Flink's memory configuration.
Although there is already a detailed setup guide[1] and migration
guide[2] officially, the
Hi, there.
In release-1.10, the memory setup of task managers has changed a lot.
I would like to provide here a third-party tool to simulate and get
the calculation result of Flink's memory configuration.
Although there is already a detailed setup guide[1] and migration
guide[2] officially, the
Hi,
在使用Flink SQL读写Oracle JDBC表时,遇到了timestamp转换异常:
Caused by: java.lang.ClassCastException: oracle.sql.TIMESTAMP cannot be cast to
java.sql.Timestamp at
org.apache.flink.table.dataformat.DataFormatConverters$TimestampConverter.toInternalImpl(DataFormatConverters.java:860)
at
I want to do a bit different hacky PoC:
* I will write a sink, that caches the results in "JVM global" memory. Then
I will write a source, that reads this cache.
* I will launch one job, that reads from Kafka source, shuffles the data to
the desired partitioning and then sinks to that cache.
*
您好,如果想订阅user-zh,可以发送邮件到user-zh-subscr...@flink.apache.org
Best,
Yangze Guo
On Fri, Mar 27, 2020 at 4:45 PM 大数据开发面试_夏永权 wrote:
>
> 申请中文社区,不知是否成功,我非常想加入社区,非常感谢。
申请中文社区,不知是否成功,我非常想加入社区,非常感谢。
Hi,您好,在使用flink的过程中遇到如下问题,个人未能解决,所以请求您指导一下,谢谢!
1. flink cancel -s $SAVEPOINT_DIR $job_id -yid $application_id 在程序有背压的时候停不掉
The program finished with the following exception:
org.apache.flink.util.FlinkException: Could not cancel job
1f768e4ca9ad5792a4844a5d12163b73.
at
Hi Apoorv,
Sorry for the late reply, have been quite busy with backlog items the past
days.
On Fri, Mar 20, 2020 at 4:37 PM Apoorv Upadhyay <
apoorv.upadh...@razorpay.com> wrote:
> Thanks Gordon for the suggestion,
>
> I am going by this repo :
>
Hello Zhu,
The host can be resolved and there are no filewalls in the cluster, so all
ports are opened.
Regards,
Vitaliy
On Fri, Mar 27, 2020 at 8:32 AM Zhu Zhu wrote:
> Hi Vitaliy,
>
> >> *Cannot serve slot request, no ResourceManager connected*
> This is not a problem, just that the JM
Hi,
对于第二个场景,可以尝试一下fast emit:
table.exec.emit.early-fire.enabled = true
table.exec.emit.early-fire.delay = 5min
PS:
1. 这个feature并没有在官方文档里面写出来,目前应该是一个实验性质的feature
2. window加了emit之后,会由原来输出append结果变成输出retract结果
Jingsong Li 于2020年3月27日周五 下午2:51写道:
> Hi,
>
> For #1:
> 创建级联的两级window:
> - 1分钟窗口
> -
??
----
??:"Utopia"https://bigdata.cs.ut.ee/keyed-watermarks-partition-aware-watermark-generation-apache-flink
Best regards
Utopia
If you are using both the Hadoop S3 and Presto S3 filesystems, you should
use s3p:// and s3a:// to distinguish between the two.
Presto is recommended for checkpointing because the Hadoop implementation
has very high latency when creating files, and because it hits request rate
limits very
Hi,
For #1:
创建级联的两级window:
- 1分钟窗口
- 5分钟窗口,计算只是保存数据,发送明细数据结果
Best,
Jingsong Lee
Hi,
In this document
https://ci.apache.org/projects/flink/flink-docs-stable/ops/filesystems/s3.html#hadooppresto-s3-file-systems-plugins,
it mentioned that
* Presto is the recommended file system for checkpointing to S3.
Is there a reason for that? Is there some bottleneck for s3 hadoop
33 matches
Mail list logo