There are multiple versions of Jackson-Databind, and the error message indicates the jar name.

[melin]

The Jackson-Databind library was introduced into a fat jar, and the following error message was displayed. It's quite difficult to solve the problem.

Caused by: com.fasterxml.jackson.databind.JsonMappingException: Scala module 2.15.2 requires Jackson Databind version >= 2.15.0 and < 2.16.0 - Found jackson-databind version 2.14.3

[cowtowncoder]

This indicates version mismatch: you have jackson-module-scala version 2.15.2, which requires version 2.15.x of jackson-databind. But fat jar (or classpath) contains version 2.14.3.

At least this is exactly what error message says. :-)

So make sure to use same minor versions (probably 2.15.2) of all Jackson components.

[JooHyukKim]

Yes, u may use build tools (Gradle, Maven) of ur choice. They all provide analysis functionalitu

[melin]

If it's just a pure Maven & Gradle project, it's quite easy to find duplicate jars. But my scenario is a bit more complicated.
In the production environment, this issue occurred because Spark manually added many jars, causing the task to fail upon startup. By using the script at https://github.com/oldratlee/useful-scripts/blob/dev-3.x/docs/java.md#-find-in-jars, it was discovered that a flat jar contained jackson-databind. If the jar file name is provided in the error message, the problem can be quickly resolved.

[cowtowncoder]

@melin Might be worth filing an issue for Jackson Scala module (https://github.com/FasterXML/jackson-module-scala/).
Not sure exactly what info is possible to add, but I'm sure exception message could be further improved to help with issues like this.

[pjfanning]

I've seen logging frameworks that can be configured to include jar info. I, personally, am not going to spend time on 'improving' the jackson-module-scala message. If people want to create fat jars, this is one of the consequences. In my opinion, they create more problems than they solve. Mashing together carefully crafted jars and steamrolling the META-INF and other contents and worse still, they are rarely thoroughly integration tested.