flink on yarn,提交方式是per job的话,如何保证高可用?

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

flink on yarn,提交方式是per job的话,如何保证高可用?

Jim Chen
请教一下:flink on yarn,提交方式是per job的话,如何保证高可用?
Reply | Threaded
Open this post in threaded view
|

Re: flink on yarn,提交方式是per job的话,如何保证高可用?

Zhenghua Gao
JM is restarted by YARN on failure [1].

[1]
https://ci.apache.org/projects/flink/flink-docs-stable/ops/jobmanager_high_availability.html#yarn-cluster-high-availability

*Best Regards,*
*Zhenghua Gao*


On Tue, Aug 13, 2019 at 4:51 PM 陈帅 <[hidden email]> wrote:

> 请教一下:flink on yarn,提交方式是per job的话,如何保证高可用?
>
Reply | Threaded
Open this post in threaded view
|

Re: flink on yarn,提交方式是per job的话,如何保证高可用?

tison
YARN AM 包含了 Dispatcher/Resource Manager/JobManager,AM 挂掉的时候 YARN 会负责拉起来
一个相同配置的 AM。

Flink 自身的高可用,per job 模式下,CheckpointStore 基于 ZooKeeper 和 HDFS 来做,可以保证。
JobGraphStore 存在内存里,不是高可用的,但是每次新拉起来的 AM 都带有用户提交的 JobGraph 信息,
所以新拉起来的 Dispatcher 也能从中知道 JobGraph 的内容。

Best,
tison.


Zhenghua Gao <[hidden email]> 于2019年8月13日周二 下午5:02写道:

> JM is restarted by YARN on failure [1].
>
> [1]
>
> https://ci.apache.org/projects/flink/flink-docs-stable/ops/jobmanager_high_availability.html#yarn-cluster-high-availability
>
> *Best Regards,*
> *Zhenghua Gao*
>
>
> On Tue, Aug 13, 2019 at 4:51 PM 陈帅 <[hidden email]> wrote:
>
> > 请教一下:flink on yarn,提交方式是per job的话,如何保证高可用?
> >
>