9c124f2cde
* Add new configurable buffer period to create gap between mark unused and kill of segment * Changes after testing * fixes and improvements * changes after initial self review * self review changes * update sql statement that was lacking last_used * shore up some code in SqlMetadataConnector after self review * fix derby compatibility and improve testing/docs * fix checkstyle violations * Fixes post merge with master * add some unit tests to improve coverage * ignore test coverage on new UpdateTools cli tool * another attempt to ignore UpdateTables in coverage check * change column name to used_flag_last_updated * fix a method signature after column name switch * update docs spelling * Update spelling dictionary * Fixing up docs/spelling and integrating altering tasks table with my alteration code * Update NULL values for used_flag_last_updated in the background * Remove logic to allow segs with null used_flag_last_updated to be killed regardless of bufferPeriod * remove unneeded things now that the new column is automatically updated * Test new background row updater method * fix broken tests * fix create table statement * cleanup DDL formatting * Revert adding columns to entry table by default * fix compilation issues after merge with master * discovered and fixed metastore inserts that were breaking integration tests * fixup forgotten insert by using pattern of sharing now timestamp across columns * fix issue introduced by merge * fixup after merge with master * add some directions to docs in the case of segment table validation issues |
||
---|---|---|
.. | ||
docker | ||
k8s | ||
script | ||
src | ||
.gitignore | ||
README.md | ||
build_run_cluster.sh | ||
k8s_run_config_file.json | ||
pom.xml | ||
quickstart-it.json | ||
stop_cluster.sh |
README.md
Integration Testing
To run integration tests, you have to specify the druid cluster the tests should use.
Druid comes with the mvn profile integration-tests for setting up druid running in docker containers, and using that cluster to run the integration tests.
To use a druid cluster that is already running, use the mvn profile int-tests-config-file, which uses a configuration file describing the cluster.
Integration Testing Using Docker
Before starting, if you don't already have docker on your machine, install it as described on Docker installation instructions. Ensure that you have at least 4GiB of memory allocated to the docker engine. (You can verify it under Preferences > Resources > Advanced.)
Also set the DOCKER_IP
environment variable to localhost on your system, as follows:
export DOCKER_IP=127.0.0.1
Optionally, you can also set APACHE_ARCHIVE_MIRROR_HOST
to override https://archive.apache.org
host. This host is used to download archives such as hadoop and kafka during building docker images:
export APACHE_ARCHIVE_MIRROR_HOST=https://example.com/remote-generic-repo
Running tests against auto brought up Docker containers
This section describes how to start integration tests against Docker containers which will be brought up automatically by following commands. If you want to build Docker images and run tests separately, see the next section.
To run all tests from a test group using Docker and Maven run the following command:
mvn verify -P integration-tests -Dgroups=<test_group>
The list of test groups can be found at
integration-tests/src/test/java/org/apache/druid/tests/TestNGGroup.java
.
Run a single test
To run only a single test using Maven:
mvn verify -P integration-tests -Dgroups=<test_group> -Dit.test=<test_name>
Parameters:
-
Test Group: Required, as certain test tasks for setup and cleanup are based on the test group. You can find the test group for a given test as an annotation in the respective test class. A list of test groups can be found at
integration-tests/src/test/java/org/apache/druid/tests/TestNGGroup.java
. The annotation uses a string constant defined inTestNGGroup.java
, be sure to use the constant value, not name. For example, if your test has the annotation:@Test(groups = TestNGGroup.BATCH_INDEX)
then use the argument-Dgroups=batch-index
. -
Test Name: Use the fully-qualified class name. For example,
org.apache.druid.tests.indexer.ITOverwriteBatchIndexTest
. -
Add
-pl :druid-integration-tests
when running integration tests for the second time or later without changing the code of core modules in between to skip up-to-date checks for the whole module dependency tree. -
Integration tests can also be run with either Java 8 or Java 11 by adding
-Djvm.runtime=#
to themvn
command, where#
can either be 8 or 11. -
Druid's configuration (using Docker) can be overridden by providing
-Doverride.config.path=<PATH_TO_FILE>
. The file must contain one property per line, the key must start withdruid_
and the format should be snake case. Note that when bringing up Docker containers through Maven and-Doverride.config.path
is provided, additional Druid routers for security group integration test (permissive tls, no client auth tls, custom check tls) will not be started.
Debugging test runs
The integration test process is fragile and can fail for many reasons when run on your machine. Here are some suggestions.
Workround for failed builds
Sometimes the command above may fail for reasons unrelated to the changes you wish to test. In such cases, a workaround is to build the code first, then use the next section to run individual tests. To build:
mvn clean package -P integration-tests -Pskip-static-checks -Pskip-tests -T1.0C -nsu
Keep the local Maven cache fresh
As you work with issues, you may be tempted to reuse already-built jars. That only works for about 24 hours,
after which Maven will helpfully start downloading snapshot jars from an upstream repository.
This is, unfortunately, a feature of the build scripts. The -nsu
option above tries to force
Maven to only look locally for snapshot jars.
Running tests against manually brought up Docker containers
-
Build docker images.
From the source root, run the following command:
mvn clean install \ -Pintegration-tests,skip-static-checks,skip-tests \ -Ddocker.run.skip=true \ -Ddocker.build.hadoop=true
NOTE:
-Ddocker.build.hadoop=true
is optional if you don't run tests against Hadoop. -
Choose a docker-compose file to start containers.
There are a few different Docker compose yamls located in "docker" folder that could be used to start containers for different tests.
-
To start basic Druid cluster (skip this if running Druid cluster with override configs):
docker-compose -f integration-tests/docker/docker-compose.yml up
-
To start Druid cluster with override configs
OVERRIDE_ENV=<PATH_TO_ENV> docker-compose -f docker-compose.yml up
-
To start tests against Hadoop
docker-compose -f docker-compose.druid-hadoop.yml up
-
To start tests againt security group
docker-compose -f docker-compose.yml -f docker-compose.security.yml up
-
-
Run tests.
Execute the following command from root module, where
<test_name>
is the class name of a test, such as ITIndexerTest.mvn verify -P integration-tests -pl integration-tests -Ddocker.build.skip=true -Ddocker.run.skip=true -Dit.test=<test_name>
Running tests from IntelliJ
Before running tests from IntelliJ, ensure you have a Druid cluster running. Modify the tests run configurations to be the following Vm options:
-Duser.timezone=UTC
-Dfile.encoding=UTF-8
-Ddruid.test.config.dockerIp=localhost
-Ddruid.zk.service.host=localhost
-Ddruid.client.https.trustStorePath=client_tls/truststore.jks
-Ddruid.client.https.trustStorePassword=druid123
-Ddruid.client.https.keyStorePath=client_tls/client.jks
-Ddruid.client.https.certAlias=druid
-Ddruid.client.https.keyManagerPassword=druid123
-Ddruid.client.https.keyStorePassword=druid123
Run tests from the test configuration often found in the top right corner of the IntelliJ IDE. The values shown above are for the default docker compose cluster. For other clusters the values will need to be changed.
Docker Compose files
-
docker-compose.base.yml
Base file that defines all containers for integration testing
-
docker-compose.yml
Defines a Druid cluster with default configuration that is used for running integration tests.
docker-compose -f docker-compose.yml up # DRUID_INTEGRATION_TEST_GROUP - an environment variable that specifies the integration test group to run. DRUID_INTEGRATION_TEST_GROUP=batch-index docker-compose -f docker-compose.yml up
You can change the default configuration using a custom configuration file. The settings in the file will override the default settings if they conflict. They will be appended to the default configuration otherwise.
# OVERRIDE_ENV - an environment variable that specifies the custom configuration file path. OVERRIDE_ENV=./environment-configs/test-groups/prepopulated-data DRUID_INTEGRATION_TEST_GROUP=query docker-compose -f docker-compose.yml up
-
docker-compose.security.yml
Defines three additional Druid router services with permissive tls, no client auth tls, and custom check tls respectively. This is meant to be used together with docker-compose.yml and is only needed for the "security" group integration test.
docker-compose -f docker-compose.yml -f docker-compose.security.yml up
-
docker-compose.druid-hadoop.yml
For starting Apache Hadoop 3.3.6 cluster with the same setup as the Druid tutorial.
docker-compose -f docker-compose.druid-hadoop.yml up
Tips & tricks for debugging and developing integration tests
Useful mvn command flags
Flag | Description |
---|---|
-Ddocker.build.skip=true | Skip building the containers. If you do not apply any change to Druid then you skip rebuilding the containers. This can save ~4 minutes. You need to build druid containers only once, after you can skip docker build step. |
-Ddocker.run.skip=true | Skip starting docker containers. This can save ~3 minutes by skipping building and bringing up the docker containers (Druid, Kafka, Hadoop, MYSQL, zookeeper, etc). Please make sure that you actually do have these containers already running if using this flag. Additionally, please make sure that the running containers are in the same state that the setup script (run_cluster.sh) would have brought it up in. |
-Ddocker.build.hadoop=true | Build the hadoop image when either running integration tests or when building the integration test docker images without running the tests. |
-Dstart.hadoop.docker=true | Start hadoop container when you need to run IT tests that utilize local hadoop docker. |
Debugging Druid while running tests
For your convenience, Druid processes running inside Docker have been debugging enabled at following debugging ports:
Process | Remote Debugging Port |
---|---|
Router with permissive tls | 5001 |
Router with no client auth tls | 5002 |
Router with custom check tls | 5003 |
Router | 5004 |
Broker | 5005 |
Coordinator | 5006 |
Historical | 5007 |
Middlemanager | 5008 |
Overlord | 5009 |
Peons (Workers on Middlemanager) | Ephemeral port assigned by debugger (check task log for port assigned to each task) |
You can use remote debugger(such as via IntelliJ IDEA's Remote Configuration) to debug the corresponding Druid process at above port.
Running Tests Using A Quickstart Cluster
When writing integration tests, it can be helpful to test against a quickstart cluster so that you can set up remote debugging with in your developer environment. This section walks you through setting up the integration tests so that it can run against a quickstart cluster running on your development machine.
NOTE: Not all features run by default on a quickstart cluster, so it may not make sense to run the entire test suite against this configuration.
Quickstart does not run with ssl, so to trick the integration tests we specify the
*_tls_url
in the config to be the same as the http url.
Make sure you have at least 6GiB of memory available before you run the tests.
The tests rely on files in the test/resources folder to exist under the path /resources, so create a symlink to make them available:
ln -s ${DRUID_HOME}/integration-tests/src/test/resources /resources
Set the cluster config file environment variable to the quickstart config:
export CONFIG_FILE=${DRUID_HOME}/integration-tests/quickstart-it.json
The test group quickstart-compatible
has tests that have been verified to work against the quickstart cluster.
There may be more tests that work, if you find that they do, please mark it as quickstart-compatible
(TestNGGroup#QUICKSTART_COMPATIBLE) and open a PR.
If you find some integration tests do not work, look at the docker files to see what setup they do. You may need to
do similar steps to get the test to work.
Then run the tests using a command similar to:
mvn verify -P int-tests-config-file -Dit.test=<test_name>
# Run all integration tests that have been verified to work against a quickstart cluster.
mvn verify -P int-tests-config-file -Dgroups=quickstart-compatible
Running Tests Using A Configuration File for Any Cluster
Make sure that you have at least 6GiB of memory available before you run the tests.
To run tests on any druid cluster that is already running, create a configuration file:
{
"broker_host": "<broker_ip>",
"broker_port": "<broker_port>",
"router_host": "<router_ip>",
"router_port": "<router_port>",
"indexer_host": "<indexer_ip>",
"indexer_port": "<indexer_port>",
"coordinator_host": "<coordinator_ip>",
"coordinator_port": "<coordinator_port>",
"middlemanager_host": "<middle_manager_ip>",
"zookeeper_hosts": "<comma-separated list of zookeeper_ip:zookeeper_port>",
"cloud_bucket": "<(optional) cloud_bucket for test data if running cloud integration test>",
"cloud_path": "<(optional) cloud_path for test data if running cloud integration test>"
}
Set the environment variable CONFIG_FILE
to the name of the configuration file:
export CONFIG_FILE=<config file name>
To run all tests from a test group using mvn run the following command: (list of test groups can be found at integration-tests/src/test/java/org/apache/druid/tests/TestNGGroup.java)
mvn verify -P int-tests-config-file -Dgroups=<test_group>
To run only a single test using mvn run the following command:
mvn verify -P int-tests-config-file -Dit.test=<test_name>
Running a Test That Uses Cloud
The integration test that indexes from Cloud or uses Cloud as deep storage is not run as part of the integration test run discussed above. Running these tests requires the user to provide their own Cloud.
Currently, the integration test supports Amazon Kinesis, Google Cloud Storage, Amazon S3, and Microsoft Azure. These can be run by providing "kinesis-index", "kinesis-data-format", "gcs-deep-storage", "s3-deep-storage", or "azure-deep-storage" to -Dgroups for Amazon Kinesis, Google Cloud Storage, Amazon S3, and Microsoft Azure respectively. Note that only one group should be run per mvn command.
For all the Cloud Integration tests, the following will also need to be provided:
- Provide -Doverride.config.path=<PATH_TO_FILE> with your Cloud credentials/configs set. See integration-tests/docker/environment-configs/override-examples/ directory for env vars to provide for each Cloud.
For Amazon Kinesis, the following will also need to be provided:
- Provide -Ddruid.test.config.streamEndpoint=<STREAM_ENDPOINT> with the endpoint of your stream set. For example, kinesis.us-east-1.amazonaws.com
For Google Cloud Storage, Amazon S3, and Microsoft Azure, the following will also need to be provided:
- Set the bucket and path for your test data. This can be done by setting -Ddruid.test.config.cloudBucket and -Ddruid.test.config.cloudPath in the mvn command or setting "cloud_bucket" and "cloud_path" in the config file.
- Copy wikipedia_index_data1.json, wikipedia_index_data2.json, and wikipedia_index_data3.json located in integration-tests/src/test/resources/data/batch_index/json to your Cloud storage at the location set in step 1.
For Google Cloud Storage, in addition to the above, you will also have to:
- Provide -Dresource.file.dir.path=<PATH_TO_FOLDER> with folder that contains GOOGLE_APPLICATION_CREDENTIALS file
For example, to run integration test for Google Cloud Storage:
mvn verify -P integration-tests -Dgroups=gcs-deep-storage -Doverride.config.path=<PATH_TO_FILE> \
-Dresource.file.dir.path=<PATH_TO_FOLDER> -Ddruid.test.config.cloudBucket=test-bucket \
-Ddruid.test.config.cloudPath=test-data-folder/
Running a Test That Uses Hadoop
The integration test that indexes from hadoop is not run as part of the integration test run discussed above. This is because druid test clusters might not, in general, have access to hadoop. This also applies to integration test that uses Hadoop HDFS as an inputSource or as a deep storage. To run integration test that uses Hadoop, you will have to run a Hadoop cluster. This can be done in two ways:
- Run Druid Docker test clusters with Hadoop container by passing -Dstart.hadoop.docker=true to the mvn command. If you have not already built the hadoop image, you will also need to add -Ddocker.build.hadoop=true to the mvn command.
- Run your own Druid + Hadoop cluster and specified Hadoop configs in the configuration file (CONFIG_FILE).
Currently, hdfs-deep-storage and other -deep-storage integration test groups can only be run with Druid Docker test clusters by passing -Dstart.hadoop.docker=true to start Hadoop container. You will also have to provide -Doverride.config.path=<PATH_TO_FILE> with your Druid's Hadoop configs set. See integration-tests/docker/environment-configs/override-examples/hdfs directory for example. Note that if the integration test you are running also uses other cloud extension (S3, Azure, GCS), additional credentials/configs may need to be set in the same file as your Druid's Hadoop configs set.
If you are running ITHadoopIndexTest with your own Druid + Hadoop cluster, please follow the below steps:
- Copy wikipedia_index_data1.json, wikipedia_index_data2.json, and wikipedia_index_data3.json located in integration-tests/src/test/resources/data/batch_index/json to your HDFS at /batch_index/json/
- Copy batch_hadoop.data located in integration-tests/src/test/resources/data/batch_index/hadoop_tsv to your HDFS at /batch_index/hadoop_tsv/ If using the Docker-based Hadoop container, the steps above are automatically done by the integration tests.
When running the Hadoop tests, you must set -Dextra.datasource.name.suffix=''
, due to https://github.com/apache/druid/issues/9788.
Option 1: Run the test using mvn (using the bundled Docker-based Hadoop cluster and building docker images at runtime):
mvn verify -P integration-tests -Dit.test=ITHadoopIndexTest -Dstart.hadoop.docker=true -Ddocker.build.hadoop=true \
-Doverride.config.path=docker/environment-configs/override-examples/hdfs -Dextra.datasource.name.suffix=''
Option 2: Run the test using mvn (using the bundled Docker-based hadoop cluster and not building images at runtime):
mvn verify -P integration-tests -Dit.test=ITHadoopIndexTest -Dstart.hadoop.docker=true -Ddocker.build.skip=true \
-Doverride.config.path=docker/environment-configs/override-examples/hdfs -Dextra.datasource.name.suffix=''
Option 3: Run the test using mvn (using the bundled Docker-based hadoop cluster and when you have already started all containers)
mvn verify -P integration-tests -Dit.test=ITHadoopIndexTest -Ddocker.run.skip=true -Ddocker.build.skip=true \
-Doverride.config.path=docker/environment-configs/override-examples/hdfs -Dextra.datasource.name.suffix=''
Option 4: Run the test using mvn (using config file for existing Hadoop cluster):
mvn verify -P int-tests-config-file -Dit.test=ITHadoopIndexTest -Dextra.datasource.name.suffix=''
In some test environments, the machine where the tests need to be executed cannot access the outside internet, so mvn cannot be run. In that case, do the following instead of running the tests using mvn:
Compile druid and the integration tests
On a machine that can do mvn builds:
cd druid
mvn clean package
cd integration_tests
mvn dependency:copy-dependencies package
Put the compiled test code into your test cluster
Copy the integration-tests directory to the test cluster.
Set CLASSPATH
TDIR=<directory containing integration-tests>/target
VER=<version of druid you built>
export CLASSPATH=$TDIR/dependency/*:$TDIR/druid-integration-tests-$VER.jar:$TDIR/druid-integration-tests-$VER-tests.jar
Run the test
java -Duser.timezone=UTC -Dfile.encoding=UTF-8 -Ddruid.test.config.type=configFile \
-Ddruid.test.config.configFile=<pathname of configuration file> org.testng.TestNG \
-testrunfactory org.testng.DruidTestRunnerFactory -testclass org.apache.druid.tests.hadoop.ITHadoopIndexTest
Writing a New Test
What should we cover in integration tests
For every end-user functionality provided by druid we should have an integration-test verifying the correctness.
Rules to be followed while writing a new integration test
Every Integration Test must follow these rules:
- Name of the test must start with a prefix "IT"
- A test should be independent of other tests
- Tests are to be written in TestNG style (http://testng.org/doc/documentation-main.html#methods)
- If a test loads some data it is the responsibility of the test to clean up the data from the cluster
How to use Guice Dependency Injection in a test
A test can access different helper and utility classes provided by test-framework in order to access Coordinator,Broker etc.. To mark a test be able to use Guice Dependency Injection - Annotate the test class with the below annotation
@Guice(moduleFactory = DruidTestModuleFactory.class)
This will tell the test framework that the test class needs to be constructed using guice.
Helper Classes provided
- IntegrationTestingConfig - configuration of the test
- CoordinatorResourceTestClient - httpclient for coordinator endpoints
- OverlordResourceTestClient - httpclient for indexer endpoints
- QueryResourceTestClient - httpclient for broker endpoints
Static Utility classes
- RetryUtil - provides methods to retry an operation until it succeeds for configurable no. of times
- FromFileTestQueryHelper - reads queries with expected results from file and executes them and verifies the results using ResultVerifier
Refer ITIndexerTest as an example on how to use dependency Injection
Running test methods in parallel
By default, test methods in a test class will be run in sequential order one at a time. Test methods for a given test
class can be set to run in parallel (multiple test methods of each class running at the same time) by excluding
the given class/package from the "AllSerializedTests" test tag section and including it in the "AllParallelizedTests"
test tag section in integration-tests/src/test/resources/testng.xml. TestNG uses two parameters, i.e.,
thread-count
and data-provider-thread-count
, for parallel test execution, which are both set to 2 for Druid integration tests.
For test using parallel execution with data provider, you will also need to set @DataProvider(parallel = true)
on your data provider method in your test class. Note that for test using parallel execution with data provider, the test
class does not need to be in the "AllParallelizedTests" test tag section and if it is in the "AllParallelizedTests"
test tag section it will actually be run with thread-count
times data-provider-thread-count
threads.
You may want to modify those values for faster execution.
See https://testng.org/doc/documentation-main.html#parallel-running and https://testng.org/doc/documentation-main.html#parameters-dataproviders for details.
Please be mindful when adding tests to the "AllParallelizedTests" test tag that the tests can run in parallel with other tests from the same class at the same time. i.e. test does not modify/restart/stop the druid cluster or other dependency containers, test does not use excessive memory starving other concurent task, test does not modify and/or use other task, supervisor, datasource it did not create.
Limitation of Druid cluster in Travis environment
By default, integration tests are run in Travis environment on commits made to open PR. These integration test jobs are required to pass for a PR to be elligible to be merged. Here are known issues and limitations to the Druid docker cluster running in Travis machine that may cause the tests to fail:
- Number of concurrent running tasks. Although the default Druid cluster config sets the maximum number of tasks (druid.worker.capacity) to 10, the actual maximum can be lowered depending on the type of the tasks. For example, running 2 range partitioning compaction tasks with 2 subtasks each (for a total of 6 tasks) concurrently can cause the cluster to intermittently fail. This can cause the Travis job to become stuck until it timeouts (50 minutes) and/or terminates after 10 mins of not receiving new output.