Submit high version compiled code jar to low version flink cluster?

The flink cluster version is 1.8.2

The application source code needs some feature only supported in 1.9.1.  So it is compiled with flink-1.9.1 denendency and builed to a fat jar with all the flink dependencies.

What it will happen if I submit the high version builed jar to the low verion flink cluster?

Thansk,

Lei

[tison]

It possibly fails with incompatibility. Flink doesn't promise such compatibility but it MIGHT work.

[hidden email] <[hidden email]> 于2019年12月30日周一 下午3:17写道：

The flink cluster version is 1.8.2

The application source code needs some feature only supported in 1.9.1.  So it is compiled with flink-1.9.1 denendency and builed to a fat jar with all the flink dependencies.

What it will happen if I submit the high version builed jar to the low verion flink cluster?

Thansk,

Lei

[Yun Tang]

Hi Lei

It's better to use the SAME version to submit job from client side. Even the major version of Flink is the same, the compatibility has not been declared to support. There exist a known issue due to some classes missing 'serialVersionUID'. [1]

[1] https://issues.apache.org/jira/browse/FLINK-13910

Best

Yun Tang

---

From: tison <[hidden email]>
Sent: Monday, December 30, 2019 15:44
To: [hidden email] <[hidden email]>
Cc: user <[hidden email]>
Subject: Re: Submit high version compiled code jar to low version flink cluster?

 

It possibly fails with incompatibility. Flink doesn't promise such compatibility but it MIGHT work.

[hidden email] <[hidden email]> 于2019年12月30日周一 下午3:17写道：

The flink cluster version is 1.8.2

The application source code needs some feature only supported in 1.9.1.  So it is compiled with flink-1.9.1 denendency and builed to a fat jar with all the flink dependencies.

What it will happen if I submit the high version builed jar to the low verion flink cluster?

Thansk,

Lei

[Arvid Heise-3]

If you explicitly need features that are only present in Flink 1.9, chances are high that your code will fail on older versions. If it's just about syntactic sugar, a valid option is to copy the new functions in your code and use that with the old version.

However, if you refer to SQL queries then your approach would be better as it's not easy to cherry-pick specific parts. Then your only option is to try it out. 

Another orthogonal approach would be to bundle Flink as a library with your application and run it dockerized. Then all applications can use their own Flink version.

On Mon, Dec 30, 2019 at 12:08 PM Yun Tang <[hidden email]> wrote:

Hi Lei

It's better to use the SAME version to submit job from client side. Even the major version of Flink is the same, the compatibility has not been declared to support. There exist a known issue due to some classes missing 'serialVersionUID'. [1]

[1] https://issues.apache.org/jira/browse/FLINK-13910

Best

Yun Tang

---

From: tison <[hidden email]>
Sent: Monday, December 30, 2019 15:44
To: [hidden email] <[hidden email]>
Cc: user <[hidden email]>
Subject: Re: Submit high version compiled code jar to low version flink cluster?

 

It possibly fails with incompatibility. Flink doesn't promise such compatibility but it MIGHT work.

Best,

tison.

[hidden email] <[hidden email]> 于2019年12月30日周一 下午3:17写道：

The flink cluster version is 1.8.2

The application source code needs some feature only supported in 1.9.1.  So it is compiled with flink-1.9.1 denendency and builed to a fat jar with all the flink dependencies.

What it will happen if I submit the high version builed jar to the low verion flink cluster?

Thansk,

Lei