Re: [DISCUSS] What parts of the Python API should we focus on next ?

2019-12-22 Thread jincheng sun
Hi Bowen, Your suggestions are very helpful for expanding the PyFlink ecology. I also mentioned above to integrate notebooks,Jupyter and Zeppelin are both very excellent notebooks. The process of integrating Jupyter and Zeppelin also requires the support of Jupyter and Zeppelin community personne

using thin jar to replace fat jar on yarn cluster mode

2019-12-22 Thread zjfpla...@hotmail.com
Hi, Does flink on yarn support using thin jar to replace fat jar? I don't want the jar of each flink task to have hundreds of MB. I want to put all the dependent packages in a single directory,and then the size of each flink task jar will be tens of KB. __

Re: using thin jar to replace fat jar on yarn cluster mode

2019-12-22 Thread tangjunli...@huitongjy.com
Specify classpath tangjunli...@huitongjy.com From: zjfpla...@hotmail.com Date: 2019-12-23 10:09 To: user; user-zh Subject: using thin jar to replace fat jar on yarn cluster mode Hi, Does flink on yarn support using thin jar to replace fat jar? I don't want the jar of each flink

Re: Re: using thin jar to replace fat jar on yarn cluster mode

2019-12-22 Thread zjfpla...@hotmail.com
When using yarn originally, we found a problem. The jars on yarn will take precedence over the jars on the specified classpath, and sometimes yarn is shared, and the jars in the lib directory of yarn cannot be modified. When Flink runs on a yarn cluster, is the jar on the classpath of flink itse

Re: 使用flink 做维表关联

2019-12-22 Thread LakeShen
Flink l.9 SQL 中支持 HBase 作为维表,不过是没有缓存的,直接来一条,去HBase 查询一条,我们这边使用 HBase ,反正2 QPS 能够处理到。 这种场景,应该能够 Cover 一些常见的场景的。 当然,如果你们公司有其他的存储,你可以在 SQL 中自定义维表即可。具体查看 LookupableTableSource。 Best wishes, LakeShen lucas.wu 于2019年12月20日周五 下午5:37写道: > hi 大家好: > > 最近有在调研使用flink做实时数仓,但是有个问题没弄清楚,就是明细表和维度表做join的时候,

Flink On K8s, build docker image very slowly, is there some way to make it faster?

2019-12-22 Thread LakeShen
Hi community , when I run the flink task on k8s , the first thing is that to build the flink task jar to Docker Image . I find that It would spend much time to build docker image. Is there some way to makr it faster. Thank your replay.

Re: Flink On K8s, build docker image very slowly, is there some way to make it faster?

2019-12-22 Thread Xintong Song
Hi Lake, Usually building a docker image should not take much time (typically less than 2 minutes). It is probably the network issue that causes the long time for image building. Of course we will need more information (e.g., logs) to confirm that, but according to our experience pulling the base

Re: Flink On K8s, build docker image very slowly, is there some way to make it faster?

2019-12-22 Thread vino yang
Hi Lake, Can you clearly count or identify which steps are taking a long time? Best, Vino LakeShen 于2019年12月23日周一 下午2:46写道: > Hi community , when I run the flink task on k8s , the first thing is that > to build the flink task jar to > Docker Image . I find that It would spend much time to buil

Re: using thin jar to replace fat jar on yarn cluster mode

2019-12-22 Thread Rui Li
Hi, I think you can try specifying dependent jars with the -C option[1] when you submit the job, and see if that meets your needs. [1] https://ci.apache.org/projects/flink/flink-docs-stable/ops/cli.html#usage On Mon, Dec 23, 2019 at 10:09 AM zjfpla...@hotmail.com < zjfpla...@hotmail.com> wrote: