testcontainer initializationError while running a test suite - docker-compose

Question

I have multiple test classes running the same docker-compose with testcontainer.
The suite fails with initializationError although each test passes when performed separately.
Here is the relevant part of the stacktrace occuring during the second test.
./gradlew e2e:test -i
io.foo.e2e.AuthTest > initializationError FAILED
org.testcontainers.containers.ContainerLaunchException: Container startup failed
at org.testcontainers.containers.GenericContainer.doStart(GenericContainer.java:330)
at org.testcontainers.containers.GenericContainer.start(GenericContainer.java:311)
at org.testcontainers.containers.DockerComposeContainer.startAmbassadorContainers(DockerComposeContainer.java:331)
at org.testcontainers.containers.DockerComposeContainer.start(DockerComposeContainer.java:178)
at io.foo.e2e.bases.BaseE2eTest$Companion.beforeAll$e2e(BaseE2eTest.kt:62)
at io.foo.e2e.bases.BaseE2eTest.beforeAll$e2e(BaseE2eTest.kt)
...
Caused by:
org.rnorth.ducttape.RetryCountExceededException: Retry limit hit with exception
at org.rnorth.ducttape.unreliables.Unreliables.retryUntilSuccess(Unreliables.java:88)
at org.testcontainers.containers.GenericContainer.doStart(GenericContainer.java:323)
... 83 more
Caused by:
org.testcontainers.containers.ContainerLaunchException: Could not create/start container
at org.testcontainers.containers.GenericContainer.tryStart(GenericContainer.java:497)
at org.testcontainers.containers.GenericContainer.lambda$doStart$0(GenericContainer.java:325)
at org.rnorth.ducttape.unreliables.Unreliables.retryUntilSuccess(Unreliables.java:81)
... 84 more
Caused by:
org.testcontainers.containers.ContainerLaunchException: Aborting attempt to link to container btraq5fzahac_worker_1 as it is not running
at org.testcontainers.containers.GenericContainer.applyConfiguration(GenericContainer.java:779)
at org.testcontainers.containers.GenericContainer.tryStart(GenericContainer.java:359)
... 86 more
It seems to me that the second test doesn't wait for the first to shutdown previous containers.
Here the base class that all tests inherit from. It is responsible for spinning up the containers.
open class BaseE2eTest {
...
companion object {
const val A = "containera_1"
const val B = "containerb_1"
const val C = "containerc_1"
val dockerCompose: KDockerComposeContainer by lazy {
defineDockerCompose()
.withLocalCompose(true)
.withExposedService(A, 8080, Wait.forListeningPort())
.withExposedService(B, 8081)
.withExposedService(C, 5672, Wait.forListeningPort())
}
class KDockerComposeContainer(file: File) : DockerComposeContainer<KDockerComposeContainer>(file)
private fun defineDockerCompose() = KDockerComposeContainer(File("../docker-compose.yml"))
#BeforeAll
#JvmStatic
internal fun beforeAll() {
dockerCompose.start()
}
#AfterAll
#JvmStatic
internal fun afterAll() {
dockerCompose.stop()
}
}
}
docker-compose version 1.27.4, build 40524192
testcontainer 1.15.2
testcontainers:junit-jupiter:1.15.2

Answer 1

After watching this talk, I realized that my testcontainers instantiation approach with Junit5 was wrong.
Here is the working code:
#TestInstance(TestInstance.Lifecycle.PER_CLASS)
open class BaseE2eTest {
...
val A = "containera_1"
val B = "containerb_1"
val C = "containerc_1"
val dockerCompose: KDockerComposeContainer by lazy {
defineDockerCompose()
.withLocalCompose(true)
.withExposedService(A, 8080, Wait.forListeningPort())
.withExposedService(B, 8081)
.withExposedService(C, 5672, Wait.forListeningPort())
}
class KDockerComposeContainer(file: File) : DockerComposeContainer<KDockerComposeContainer>(file)
private fun defineDockerCompose() = KDockerComposeContainer(File("../docker-compose.yml"))
#BeforeAll
fun beforeAll() {
dockerCompose.start()
}
#AfterAll
fun afterAll() {
dockerCompose.stop()
}
}
Now the test suite passes.

Answer 2

For anyone stumbling upon this, I had a similar problem that suddenly occurred after months of working perfectly fine. It was caused by this problem: Docker "ERROR: could not find an available, non-overlapping IPv4 address pool among the defaults to assign to the network"
After removing all unused networks the problem was fixed