ComposeContainer mount different file and restart init container

[bilak]

Hello,
I'm using ComposeContainer to spin up some application-container (e.g. Keycloak) and then configure it by let's say init-container (keycloak-config-cli). When init-container finishes it works, it just stops with correct exit code.

example of docker-compose.yml

services:
  keycloak:
    image: quay.io/keycloak/keycloak:${KEYCLOAK_VERSION}
    command: >
      start-dev --log-level="org.keycloak:debug,org.keycloak.transaction:info,org.keycloak.services.scheduled:info"
  keycloak-config-cli:
    image: quay.io/adorsys/keycloak-config-cli:${KEYCLOAK_CONFIG_CLI_VERSION}  
    depends_on:
      - keycloak
    volumes:
      - "./config/realms:/config"

I'd like to reuse the initial state of application-container and just do some changes in init-container (e.g. mount different config files) and then start it again to do a re-configuration stuff for application-container. Is that possible with current implementation? Any hints how to do that?

thanks

[MILLERMARRU]

ComposeContainer does not support restarting individual services or hot-swapping mounted files mid-test — it manages the compose lifecycle as a single unit. However, there are two practical approaches for your Keycloak + keycloak-config-cli scenario:

Approach 1 — Multiple compose override files, one container per test class

Create a base docker-compose.yml plus per-scenario overrides, and spin up a fresh ComposeContainer per test class (Testcontainers reuses the container across tests within the same class):

src/test/resources/
  docker-compose.yml           ← base: keycloak service
  docker-compose-realm-a.yml   ← override: config-cli with realm-a config
  docker-compose-realm-b.yml   ← override: config-cli with realm-b config

// Test class A
@Testcontainers
class RealmATest {

    @Container
    static ComposeContainer env = new ComposeContainer(
            new File("src/test/resources/docker-compose.yml"),
            new File("src/test/resources/docker-compose-realm-a.yml"))
        .withExposedService("keycloak", 8080, Wait.forHttp("/health/ready").forStatusCode(200));

    // all tests in this class share the same container
}

// Test class B uses a separate container with its own config
@Testcontainers
class RealmBTest {

    @Container
    static ComposeContainer env = new ComposeContainer(
            new File("src/test/resources/docker-compose.yml"),
            new File("src/test/resources/docker-compose-realm-b.yml"))
        .withExposedService("keycloak", 8080, Wait.forHttp("/health/ready").forStatusCode(200));
}

Approach 2 — Apply Keycloak config via REST API between tests (no restart needed)

Instead of remounting files and restarting keycloak-config-cli, push realm configuration directly through Keycloak's Admin REST API after the container is up. This is faster and avoids any restart:

@BeforeEach
void applyRealmConfig() throws Exception {
    String adminToken = getAdminToken(); // POST /realms/master/protocol/openid-connect/token

    // Import realm JSON via Admin API
    String realmJson = Files.readString(Path.of("src/test/resources/realm-test.json"));
    HttpClient.newHttpClient().send(
        HttpRequest.newBuilder()
            .uri(URI.create(keycloakUrl + "/admin/realms"))
            .header("Authorization", "Bearer " + adminToken)
            .header("Content-Type", "application/json")
            .POST(HttpRequest.BodyPublishers.ofString(realmJson))
            .build(),
        HttpResponse.BodyHandlers.discarding()
    );
}

@AfterEach
void cleanupRealm() throws Exception {
    // DELETE /admin/realms/{realmName} to reset between tests
    String adminToken = getAdminToken();
    HttpClient.newHttpClient().send(
        HttpRequest.newBuilder()
            .uri(URI.create(keycloakUrl + "/admin/realms/test-realm"))
            .header("Authorization", "Bearer " + adminToken)
            .DELETE()
            .build(),
        HttpResponse.BodyHandlers.discarding()
    );
}

This approach keeps Keycloak running continuously and only changes realm state — no container restart, no volume remounting.

Why Approach 2 is preferred

• Much faster: realm import via API takes ~200ms vs. container restart (~5-10s)
• No Testcontainers limitation to work around
• Cleaner isolation: each test gets a fresh realm deleted after the test
• Works with the existing keycloak-config-cli workflow — just run the same JSON through the REST API instead of mounting it