QuarkusTest starts all dev services for all test profiles at startup, causing resource strain even when only a subset of tests are run

[charlesjiang-greengate]

Describe the bug

With Quarkus 3.22.x and the new test classloading rewrite (see blog post), QuarkusTest now starts a separate instance of each required dev service for every test profile at startup. This happens regardless of which tests are actually selected to run (e.g., via test tags or filters). As a result, all dev services for all profiles are started up front, even if only a subset of tests will execute. This can quickly exhaust resources on CI servers, especially when multiple heavy dev services (e.g., LocalStack) are involved.

Expected behavior

Only the dev services required for the tests that will actually be executed should be started. If test selection (e.g., by tag or class) means some profiles/tests will not run, their dev services should not be started. There should be a way to toggle between the new and previous behavior to optimize for CI resource usage.

Actual behavior

All dev services for all test profiles are started at the beginning of the test run, regardless of which tests are actually selected to run. This leads to unnecessary resource consumption and potential port conflicts, even if only a single test is executed.

Example output from docker ps -a (showing multiple LocalStack and Ryuk containers started for different profiles):

docker ps -a
CONTAINER ID   IMAGE                                       COMMAND                  CREATED          STATUS                           PORTS                                              NAMES
cfb993acd253   localstack/localstack:4.2.0                 "sh -c 'while [ ! -f…"   1 second ago     Up 1 second (health: starting)   4510-4559/tcp, 5678/tcp, 0.0.0.0:60644->4566/tcp   practical_vaughan
ae8051b47af5   testcontainers/ryuk:0.11.0                  "/bin/ryuk"              2 seconds ago    Up 2 seconds                     0.0.0.0:60630->8080/tcp                            testcontainers-ryuk-3a968fbf-f9c8-4d59-a50d-530524dfd396
e1e06af6ddef   localstack/localstack:4.2.0                 "sh -c 'while [ ! -f…"   11 seconds ago   Up 11 seconds (healthy)          4510-4559/tcp, 5678/tcp, 0.0.0.0:60550->4566/tcp   distracted_vaughan
b24635eecb86   testcontainers/ryuk:0.11.0                  "/bin/ryuk"              12 seconds ago   Up 12 seconds                    0.0.0.0:60540->8080/tcp                            testcontainers-ryuk-455b5c1f-4138-4d1d-8afa-a8186510aa32
cd8c80fa2a1c   localstack/localstack:4.2.0                 "sh -c 'while [ ! -f…"   22 seconds ago   Up 22 seconds (healthy)          4510-4559/tcp, 5678/tcp, 0.0.0.0:60457->4566/tcp   relaxed_hoover
eb11dab071b0   testcontainers/ryuk:0.11.0                  "/bin/ryuk"              23 seconds ago   Up 23 seconds                    0.0.0.0:60449->8080/tcp                            testcontainers-ryuk-302e8408-7dbb-4669-8abf-cfd982c86232
167b83a3c4b1   postgres:17                                 "docker-entrypoint.s…"   7 hours ago      Up 7 hours                       0.0.0.0:56484->5432/tcp                            charming_robinson

How to Reproduce?

1. Create a Quarkus project with multiple test profiles, each requiring a different dev service (e.g., LocalStack, Postgres).
2. Annotate test classes with @QuarkusTest and @TestProfile(...) for each profile.
3. Run tests with a filter (e.g., by tag or class) so that only a subset of tests should execute. (./mvnw clean compile test -Dgroups=test1
4. Observe that all dev services for all profiles are started at the beginning of the test run, regardless of which tests are actually executed.

See

code-with-quarkus.zip

Output of uname -a or ver

Darwin MacBook-Pro-2.local 24.4.0 Darwin Kernel Version 24.4.0: Fri Apr 11 18:33:47 PDT 2025; root:xnu-11417.101.15~117/RELEASE_ARM64_T6030 arm64

Output of java -version

openjdk version "21.0.4" 2024-07-16 LTS OpenJDK Runtime Environment Corretto-21.0.4.7.1 (build 21.0.4+7-LTS) OpenJDK 64-Bit Server VM Corretto-21.0.4.7.1 (build 21.0.4+7-LTS, mixed mode, sharing)

Quarkus version or git rev

3.22.2

Build tool (ie. output of mvnw --version or gradlew --version)

Apache Maven 3.9.9 (8e8579a9e76f7d015ee5ec7bfcdc97d260186937) Maven home: /Users/charlesjiang/.m2/wrapper/dists/apache-maven-3.9.9-bin/33b4b2b4/apache-maven-3.9.9 Java version: 21.0.4, vendor: Amazon.com Inc., runtime: /Library/Java/JavaVirtualMachines/amazon-corretto-21.jdk/Contents/Home Default locale: en_NZ, platform encoding: UTF-8 OS name: "mac os x", version: "15.4.1", arch: "x86_64", family: "mac"

Additional information

• This behavior can cause CI jobs to fail due to resource exhaustion.
• There is currently no way to revert to the previous behavior or to defer dev service startup until it is known which tests will run.
• A toggle or smarter dev service startup logic would help optimize resource usage, especially in CI environments.

[quarkus-bot]

/cc @geoand (devservices)

[geoand]

cc @holly-cummins

[cvgaviao]

I don't know if it is related.
But I tried now to run a build on Jenkins: mvn clean test -Dquarkus.profile=jenkins. That profile has a datasource configured.
But the test is failing due to a missing Docker.

Seems that the profile is not being considered.

But the profile is working well when I run with dev: mvn clean quarkus:dev -Dquarkus.profile=jenkins

[holly-cummins]

This will be resolved by #45785, so I'm going to close this as a duplicate of that item. The description of #45785 talks more about port conflicts rather than unnecessary resource usage, but the basic idea of "let's not start the dev service until we actually want it" is the same in both scenarios.

@cvgaviao, do you mind raising a new issue for what you're seeing? I think it's a different issue. It's not obvious to me why the behaviour would be different in dev mode and normal mode, because most of the code paths involved in profile detection and dev service startup are exactly the same, but that's what debuggers are for :)

[cvgaviao]

@holly-cummins , I have reproduced it at home, and created the issue #47782

[holly-cummins]

@holly-cummins , I have reproduced it at home, and created the issue #47782

Thanks!