Hi everybody, we have a Gradle 8.11.1 multi-module project. After I turned on the configuration cache, the build is running tests from different modules in parallel. This is not working for us as each module need to starts docker containers. In parallel the container are fighting for the same network port which breaks tests. I tried to disable parallel execution with

org.gradle.parallel=false

in

gradle.properties

, but the project still builds in parallel. Only after I turned off the the configuration cache, the build is sequential again. Do you have any ideas what is going on?

https://docs.gradle.org/current/userguide/configuration_cache.html#config_cache:intro:performance_improvements

All tasks run in parallel by default, subject to dependency constraints.

With configuration cache tasks always run in parallel,

org.gradle.parallel

has no effect. In order to limit parallelism, you can use a build service with

maxParallelUsages = 1

. You can then connect it to test tasks with

Task.usesService()

. Its implementation can be empty.

you could potentially use

--max-workers=1

but that will slow your whole build down

@Mikhail Lopatkin thanks, using the service is probably the simplest solution for me. @ephemient, I don't think docker networks can help me in my case. Tests are running in Gradle on the host and containers are just the infra like Postges.

Please be aware that for a multi-module project it is better to register the build service in a settings plugin or in the root project, to avoid classpath shenanigans.

Setting "org.gradle.workers.max=1" turned out to be even easier for us as our project wasn't build in parallel previously. But I still want to play around with the the services because not all our tests set require the shared resources. So some stuff can really run in parallel.

I don't think docker networks can help me in my case. Tests are running in Gradle on the host and containers are just the infra like Postges.

No, but you probably could use different (dynamic) ports so they can run in parallel. (ideally let Docker pick a dynamic port, i.e. use

-p 5432

rather than

-p 127.0.0.1:5432:5432

if using the command line, then ask docker for the port it picked, e.g. using `docker inspect`: https://docs.docker.com/reference/cli/docker/inspect/#list-all-port-bindings)

Guys, there is probably a solution how to solve the issue of creating multiple non-conflicting infras but they is not my concern right now. Thanks for trying to help.