T只是时间格式显示问题, 数据格式都是timestamp(3), 这个和T应该无关的
在 2021-08-16 13:45:12,"Geoff nie" 写道:
>谢谢你!第二个问题确实是我版本太低问题,我flink版本是1.12.1。
>第一个问题,是因为我通过flink写入iceberg
>表中,然后通过presto查询iceberg表,其他字段的表都可以查询,但是当写入的是含有TIMESTAMP 类型的表时,presto查询如下报错:
>
>Query failed (#20210816_020321_00011_wa8bs) in your-presto: Cannot
Hi!
Thanks for your interest in contributing to Flink. Currently most of the
committers are busy with the upcoming Flink 1.14 so there might be few
people having their eyes on the new PRs, especially if they do not exist in
a JIRA issue.
Please follow Jing Zhang's advice by first creating the
Hi!
As you mentioned that the configuration fetching is very infrequent, why
don't you use a blocking approach to send HTTP requests and receive
responses? This seems like a more reasonable solution to me.
Rion Williams 于2021年8月17日周二 上午4:00写道:
> Hi all,
>
> I've been exploring a few different
Hi Camile,
First of all, thanks for the great contribution, the document improvement
is very helpful.
but I don't know why nobody merges it and no comment.
>
Maybe we could try the following way to start the first contribution,
please go document [1] for more detailed information.
1. Please make
Hi Joe,
>
> caused by: org.apache.calcite.runtime.CalciteContextException: From line
> 1, column 139 to line 1, column 181: Cannot apply '$TUMBLE' to arguments of
> type '$TUMBLE(, )'. Supported form(s):
> '$TUMBLE(, )'
The first parameter of Group Window Function [1] must be a field with time
退订
您好:Flink程序部署在Yran上以Appliation Mode 模式启动的,在没有采用HA
模式的时候可以正常启动,配置了HA之后,启动异常,麻烦帮忙看下是什么原因导致的.
HA 配置如下:
high-availability: zookeeper high-availability.storageDir:
hdfs://mycluster/flink/ha high-availability.zookeeper.quorum:
zk-1:2181,zk-2:2181,zk-3:2181 high-availability.zookeeper.path.root:
Hi, all
I'm a Flink user. recently I find some problems when I use Flink, it
takes some time to understand the internal mechanisms. This really makes me
know more about Flink, but I think the doc can be clearer, so I open some
merge requests for the doc:
-
By the way, this post in Chinese showed how we do it exactly with code.
https://yiksanchan.com/posts/flink-bulk-insert-redis
And yes it had buffered writes support by leveraging Flink operator state,
and Redis Pipelining. Feel free to let you know if you have any questions.
On Tue, Aug 17,
Hi Jin,
I was in the same shoes. I tried bahir redis connector at first, then I
felt it was very limited, so I rolled out my own. It was actually quite
straightforward.
All you need to do is to extend RichSinkFunction, then put your logic
inside. Regarding Redis clients, Jedis
Hi, Jin
IIUC, the DataStream connector `RedisSink` can still be used. However,
the Table API connector `RedisTableSink` might not work (at least in
the future) because it is implemented based on the deprecated Table
connector abstraction. You can still give it a try, though.
Best,
Yangze Guo
On
Hi, Abhishek,
Do you see something like "Fatal error occurred while executing the
TaskManager" in your log or would you like to provide the whole task
manager log?
Best,
Yangze Guo
On Tue, Aug 17, 2021 at 5:17 AM Abhishek Rai wrote:
>
> Hello,
>
> In our production environment, running Flink
is apache bahir still a thing? it hasn't been touched for months (since
redis 2.8.5).
as such, looking at the current flink connector docs, it's no longer
pointing to anything from the bahir project. looking around in either the
flink or bahir newsgroups doesn't turn up anything regarding
I will definitely consider the contribution idea :)
On Mon, Aug 16, 2021 at 3:16 PM David Morávek wrote:
> Hi Rob,
>
> there is currently no on-going effort for this topic, I think this would
> be a really great contribution though. This seems to be pushing RabbitMQ
> towards new usages ;)
>
>
Hello,
In our production environment, running Flink 1.13 (Scala 2.11), where Flink
has been working without issues with a dozen or so jobs running for a
while, Flink taskmanager started crash looping with a period of ~4 minutes
per crash. The stack trace is not very informative, therefore
Hi all,
I've been exploring a few different options for storing tenant-specific
configurations within Flink state based on the messages I have flowing
through my job. Initially I had considered creating a source that would
periodically poll an HTTP API and connect that stream to my original event
Hi Rob,
there is currently no on-going effort for this topic, I think this would be
a really great contribution though. This seems to be pushing RabbitMQ
towards new usages ;)
Best,
D.
On Mon, Aug 16, 2021 at 8:16 PM Rob Englander
wrote:
> I'm wondering if there's any work underway to develop
Hi David,
You’re correct about classpathing problems – thanks for your help in spotting
them. I was able to get past that exception by removing some conflicting
packages in my shaded JAR, but I’m seeing something else that’s interesting.
With the 2 threads trying to submit jobs, one of the
I'm wondering if there's any work underway to develop DataSource/DataSink
for RabbitMQ's streams recently released in RMQ 3.9?
Rob Englander
Hello,
I am trying to configure TLS communication for a Flink cluster running on
Kubernetes. I am currently using the BCFKS format and setting that as default
via javax.net.ssl.keystoretype and javax.net.ssl.truststoretype (which are
injected in the environment variable FLINK_ENV_JAVA_OPTS).
Hi all,
I am on Flink 1.12.3.
So here’s the scenario: I have a Kafka topic as a source, where each record repsents a change to an append only audit log. The kafka record has the following fields:
id (unique identifier for that audit log entry)
operation id (is shared across
你是指这个吗?
CREATE [TEMPORARY|TEMPORARY SYSTEM] FUNCTION
[IF NOT EXISTS] [catalog_name.][db_name.]function_name
AS identifier [LANGUAGE JAVA|SCALA|PYTHON]
使用SQL注册UDF
详情见:
https://ci.apache.org/projects/flink/flink-docs-release-1.13/docs/dev/table/sql/create/#create-function
Ada Luna
Thanks for this suggestion David, it's extremely helpful.
Since this will vary depending on the elements retrieved from a separate
stream, I'm guessing something like the following would be roughly the
avenue to continue down:
fun main(args: Array) {
val parameters =
Great, let me know if that helped ;)
Best,
D.
On Mon, Aug 16, 2021 at 4:36 PM László Ciople
wrote:
> The events are json dictionaries and the key is a field which represents a
> device id, or if it doesn't exist, then actually a *hashCode *of the
> device object in the dictionary is used. So
My intuition is that you have a non-deterministic shuffle key. If you
perform any "per-key" operation, you need to make sure that the same key
always end up in the same partition. To simplify this, it means that the
key needs to have a consistent *hashCode* and *equals* across different
JVMs.
Hi Prasanna,
here are some quick thoughts
1) Batching is an aggregation operation.But what I have seen in the
> examples of windowing is that they get the count/max/min operation in the
> particular window. So could the batching be implemented via a windowing
> mechanism ?
>
I guess a custom
Hello,
I am trying to write a Flink application which receives data from Kafka,
does processing on keyed windowed streams and sends results on a
different topic.
Shortly after the job is started it fails with a NullPointerException in
StateTable.put(). I am really confused by this error, because I
Hi Vamshi,
>From your configuration I'm guessing that you're using Amazon S3 (not any
implementation such as Minio).
Two comments:
- *s3.endpoint* should not contain bucket (this is included in your s3
path, eg. *s3:///*)
- "*s3.path.style.access*: true" is only correct for 3rd party
Hi Andreas,
Per-job and session deployment modes should not be affected by this FLIP.
Application mode is just a new deployment mode (where job driver runs
embedded within JM), that co-exists with these two.
>From information you've provided, I'd say your actual problem is this
exception:
```
Hi Piotr,
unfortunately this is a known long-standing issue [1]. The problem is that
ORC format is not using Flink's filesystem abstraction for actual reading
of the underlying file, so you have to adjust your Hadoop config
accordingly. There is also a corresponding SO question [2] covering this.
Thanks, I'll check more about job tuning.
On Mon, 16 Aug 2021 at 06:28, Caizhi Weng wrote:
> Hi!
>
> if I use parallelism of 2 or 4 - it takes the same time.
>>
> It might be that there is no data in some parallelisms. You can click on
> the nodes in Flink web UI and see if it is the case for
taskmanager.network.memory.buffers-per-channel
把这个参数从默认的2调整成5,反压的PoolUsage就和网上的文章一致了,这是为什么?
Ada Luna 于2021年8月16日周一 下午4:17写道:
>
> 在网上看文章一般反压源头的inputPoolUsage应该是高的,其他被反压算子的inputPoolUsage也应该是高的。但是我最近发现的反压inputPoolUsage全是空,是Flink的反压机制就是这样,还是说这个版本的Metrics有问题。
>
> Ada Luna 于2021年8月16日周一 下午4:16写道:
> >
Hi Pranjul,
which deployment mode are you using?
- For session cluster, I'd say it's possible that memory grows with # of
jobs.
- For application mode, there is actual user-code executed, so if you're
using some native libraries in your job driver, that may be another reason
for the growing
在网上看文章一般反压源头的inputPoolUsage应该是高的,其他被反压算子的inputPoolUsage也应该是高的。但是我最近发现的反压inputPoolUsage全是空,是Flink的反压机制就是这样,还是说这个版本的Metrics有问题。
Ada Luna 于2021年8月16日周一 下午4:16写道:
>
> 版本1.10.1
> 最近我观察很多FlinkSQL 任务的反压指标发现,反压为High算子的outputPoolUsage是满的
> inputPoolUsage是空,反压源头inputPoolUsage和outputPoolUsage都是空的,这是正常的嘛。
版本1.10.1
最近我观察很多FlinkSQL 任务的反压指标发现,反压为High算子的outputPoolUsage是满的
inputPoolUsage是空,反压源头inputPoolUsage和outputPoolUsage都是空的,这是正常的嘛。
Hi Pranjul,
First of all, you adopted on-heap state backend: HashMapStateBackend, which
would not use native off-heap memory. Moreover, JobManager would not initialize
any keyed state backend instance. And if not enable high availability,
JobManagerCheckpointStorage would also not use direct
https://www.yuque.com/jeffzhangjianfeng/gldg8w/dthfu2
Ada Luna 于2021年8月16日周一 下午2:26写道:
> 目前注册UDF要通过Table API。
> 未来会通过SQL直接将UDF注册到上下文中吗?
>
--
Best Regards
Jeff Zhang
目前注册UDF要通过Table API。
未来会通过SQL直接将UDF注册到上下文中吗?
38 matches
Mail list logo