Waiting for containers to start or be ready
Wait strategies vs Startup strategies
Wait strategy: is the container in a state that is useful for testing. This is generally approximated as 'can we talk to this container over the network'. However, there are quite a few variations and nuances.
Startup strategy: did a container reach the desired running state. Almost always this just means 'wait until the container is running' - for a daemon process in a container this is the goal. Sometimes we need to wait until the container reaches a running state and then exits - this is the 'one shot startup' strategy, only used for cases where we need to run a one off command in a container but not a daemon.
Wait Strategies
Ordinarily Testcontainers will wait for up to 60 seconds for the container's first mapped network port to start listening.
This simple measure provides a basic check whether a container is ready for use.
public GenericContainer nginx = new GenericContainer(DockerImageName.parse("nginx:1.9.4")) //
.withExposedPorts(80);
If the default 60s timeout is not sufficient, it can be altered with the withStartupTimeout()
method.
If waiting for a listening TCP port is not sufficient to establish whether the container is ready, you can use the
waitingFor()
method with other WaitStrategy
implementations as shown below.
HTTP Wait strategy examples
You can choose to wait for an HTTP(S) endpoint to return a particular status code.
Waiting for 200 OK
public GenericContainer nginxWithHttpWait = new GenericContainer(DockerImageName.parse("nginx:1.9.4"))
.withExposedPorts(80)
.waitingFor(Wait.forHttp("/"));
Variations on the HTTP wait strategy are supported, including:
Waiting for multiple possible status codes
Wait.forHttp("/")
.forStatusCode(200)
.forStatusCode(301);
Waiting for a status code that matches a predicate
Wait.forHttp("/all")
.forStatusCodeMatching(it -> it >= 200 && it < 300 || it == 401);
Using TLS
Wait.forHttp("/all")
.usingTls();
Healthcheck Wait strategy examples
If the used image supports Docker's Healthcheck feature, you can directly leverage the healthy
state of the container as your wait condition:
Wait.forHealthcheck();
Log output Wait Strategy
In some situations a container's log output is a simple way to determine if it is ready or not. For example, we can wait for a `Ready' message in the container's logs as follows:
public GenericContainer containerWithLogWait = new GenericContainer(DockerImageName.parse("redis:5.0.3"))
.withExposedPorts(6379)
.waitingFor(Wait.forLogMessage(".*Ready to accept connections.*\\n", 1));
Other Wait Strategies
For further options, check out the Wait
convenience class, or the various subclasses of WaitStrategy
.
If none of these options meet your requirements, you can create your own subclass of
AbstractWaitStrategy
with an
appropriate wait mechanism in waitUntilReady()
.
The GenericContainer.waitingFor()
method accepts any valid WaitStrategy
.
Startup check Strategies
Ordinarily Testcontainers will check that the container has reached the running state and has not exited. In order to do that inspect is executed against the container and state parameter is extracted.
All logic is implemented in StartupCheckStrategy
child classes.
Running startup strategy example
This is the strategy used by default. Testcontainers just checks if container is running.
Implemented in IsRunningStartupCheckStrategy
class.
One shot startup strategy example
This strategy is intended for use with containers that only run briefly and exit of their own accord. As such, success is deemed to be when the container has stopped with exit code 0.
public GenericContainer<?> bboxWithOneShot = new GenericContainer<>(DockerImageName.parse("busybox:1.31.1"))
.withCommand(String.format("echo %s", HELLO_TESTCONTAINERS))
.withStartupCheckStrategy(
new OneShotStartupCheckStrategy().withTimeout(Duration.ofSeconds(3))
);
Indefinite one shot startup strategy example
Variant of one shot strategy that does not impose a timeout. Intended for situation such as when a long running task forms part of container startup.
It has to be assumed that the container will stop of its own accord, either with a success or failure exit code.
public GenericContainer<?> bboxWithIndefiniteOneShot = new GenericContainer<>(
DockerImageName.parse("busybox:1.31.1")
)
.withCommand("sh", "-c", String.format("sleep 5 && echo \"%s\"", HELLO_TESTCONTAINERS))
.withStartupCheckStrategy(
new IndefiniteWaitOneShotStartupCheckStrategy()
);
Minimum duration startup strategy example
Checks that the container is running and has been running for a defined minimum period of time.
public GenericContainer<?> bboxWithMinimumDuration = new GenericContainer<>(
DockerImageName.parse("busybox:1.31.1")
)
.withCommand("sh", "-c", String.format("sleep 5 && echo \"%s\"", HELLO_TESTCONTAINERS))
.withStartupCheckStrategy(
new MinimumDurationRunningStartupCheckStrategy(Duration.ofSeconds(1))
);
Other startup strategies
If none of these options meet your requirements, you can create your own subclass of
StartupCheckStrategy
with an appropriate startup check mechanism in waitUntilStartupSuccessful()
.
Or you can leave it as is and just implement the checkStartupState(DockerClient dockerClient, String containerId)
if you still want to check state
periodically.
Depending on another container
Sometimes, a container relies on another container to be ready before it should start itself. An example of this might be a database that needs to be started before your application container can link to it. You can tell a container that it depends on another container by using the dependsOn
method:
public GenericContainer<?> redis = new GenericContainer<>("redis:3.0.2").withExposedPorts(6379);
@Rule
public GenericContainer<?> nginx = new GenericContainer<>("nginx:1.9.4").dependsOn(redis).withExposedPorts(80);