This module includes both unit tests, which can run in isolation without connecting to the Azure Storage service, and integration tests, which require a working connection to interact with a container. Unit test suites follow the naming convention Test*.java. Integration tests follow the naming convention ITest*.java.
The Apache Jenkins infrastucture does not run any cloud integration tests, due to the need to keep credentials secure.
This is important: patches which do not include this declaration will be ignored
This policy has proven to be the only mechanism to guarantee full regression testing of code changes. Why the declaration of region? Two reasons
You don’t need to test from a VM within the Azure infrastructure, all you need are credentials.
It’s neither hard nor expensive to run the tests; if you can’t, there’s no guarantee your patch works. The reviewers have enough to do, and don’t have the time to do these tests, especially as every failure will simply make for a slow iterative development.
Please: run the tests. And if you don’t, we are sorry for declining your patch, but we have to.
Some of the tests do fail intermittently, especially in parallel runs. If this happens, try to run the test on its own to see if the test succeeds.
If it still fails, include this fact in your declaration. We know some tests are intermittently unreliable.
The tests are designed to be configurable for different timeouts. If you are seeing problems and this configuration isn’t working, that’s a sign of the configuration mechanism isn’t complete. If it’s happening in the production code, that could be a sign of a problem which may surface over long-haul connections. Please help us identify and fix these problems — especially as you are the one best placed to verify the fixes work.
The hadoop-azure module includes a full suite of unit tests. Many of the tests will run without additional configuration by running mvn test. This includes tests against mocked storage, which is an in-memory emulation of Azure Storage.
The integration tests are designed to test directly against an Azure storage service, and require an account and credentials in order to run.
This is done by creating the file to src/test/resources/azure-auth-keys.xml and setting the name of the storage account and its access key.
For example:
<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<configuration>
<property>
<name>fs.azure.wasb.account.name</name>
<value>{ACCOUNTNAME}.blob.core.windows.net</value>
</property>
<property>
<name>fs.azure.account.key.{ACCOUNTNAME}.blob.core.windows.net</name>
<value>{ACCOUNT ACCESS KEY}</value>
</property>
</configuration>
To run contract tests, set the WASB file system URI in src/test/resources/azure-auth-keys.xml and the account access key. For example:
<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<configuration>
<property>
<name>fs.contract.test.fs.wasb</name>
<value>wasb://{CONTAINERNAME}@{ACCOUNTNAME}.blob.core.windows.net</value>
<description>The name of the azure file system for testing.</description>
</property>
<property>
<name>fs.azure.account.key.{ACCOUNTNAME}.blob.core.windows.net</name>
<value>{ACCOUNT ACCESS KEY}</value>
</property>
</configuration>
Overall, to run all the tests using mvn test, a sample azure-auth-keys.xml is like following:
<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<configuration>
<property>
<name>fs.azure.wasb.account.name</name>
<value>{ACCOUNTNAME}.blob.core.windows.net</value>
</property>
<property>
<name>fs.azure.account.key.{ACCOUNTNAME}.blob.core.windows.net</name>
<value>{ACCOUNT ACCESS KEY}</value>
</property>
<property>
<name>fs.contract.test.fs.wasb</name>
<value>wasb://{CONTAINERNAME}@{ACCOUNTNAME}.blob.core.windows.net</value>
</property>
</configuration>
DO NOT ADD azure-auth-keys.xml TO REVISION CONTROL. The keys to your Azure Storage account are a secret and must not be shared.
After completing the configuration, execute the test run through Maven.
mvn -T 1C clean verify
It’s also possible to execute multiple test suites in parallel by passing the parallel-tests property on the command line. The tests spend most of their time blocked on network I/O, so running in parallel tends to complete full test runs faster.
mvn -T 1C -Dparallel-tests clean verify
Some tests must run with exclusive access to the storage container, so even with the parallel-tests property, several test suites will run in serial in a separate Maven execution step after the parallel tests.
By default, parallel-tests runs 4 test suites concurrently. This can be tuned by passing the testsThreadCount property.
mvn -T 1C -Dparallel-tests -DtestsThreadCount=8 clean verify
mvn -T 1C clean test mvn -T 1C -Dparallel-tests clean test mvn -T 1C -Dparallel-tests -DtestsThreadCount=8 clean test
To run only a specific named subset of tests, pass the test property for unit tests or the it.test property for integration tests.
mvn -T 1C clean test -Dtest=TestRollingWindowAverage mvn -T 1C clean verify -Dscale -Dit.test=ITestFileSystemOperationExceptionMessage -Dtest=none mvn -T 1C clean verify -Dtest=none -Dit.test=ITest*
Note
When running a specific subset of tests, the patterns passed in test and it.test override the configuration of which tests need to run in isolation in a separate serial phase (mentioned above). This can cause unpredictable results, so the recommendation is to avoid passing parallel-tests in combination with test or it.test. If you know that you are specifying only tests that can run safely in parallel, then it will work. For wide patterns, like ITest* shown above, it may cause unpredictable test failures.
The command line shell may try to expand the “*” and sometimes the “#” symbols in test patterns. In such situations, escape the character it with a “\” prefix. Example:
mvn -T 1C clean verify -Dtest=none -Dit.test=ITest\*
Integration test results and logs are stored in target/failsafe-reports/. An HTML report can be generated during site generation, or with the surefire-report plugin:
# for the unit tests mvn -T 1C surefire-report:report-only # for the integration tests mvn -T 1C surefire-report:failsafe-report-only # all reports for this module mvn -T 1C site:site
There are a set of tests designed to measure the scalability and performance at scale of the filesystem client, Scale Tests. Tests include: creating and traversing directory trees, uploading large files, renaming them, deleting them, seeking through the files, performing random IO, and others. This makes them a foundational part of the benchmarking.
By their very nature they are slow. And, as their execution time is often limited by bandwidth between the computer running the tests and the Azure endpoint, parallel execution does not speed these tests up.
The tests are enabled if the scale property is set in the maven build this can be done regardless of whether or not the parallel test profile is used
mvn -T 1C verify -Dscale mvn -T 1C verify -Dparallel-tests -Dscale -DtestsThreadCount=8
The most bandwidth intensive tests (those which upload data) always run sequentially; those which are slow due to HTTPS setup costs or server-side actions are included in the set of parallelized tests.
Some of the tests can be tuned from the maven build or from the configuration file used to run the tests.
mvn -T 1C verify -Dparallel-tests -Dscale -DtestsThreadCount=8 -Dfs.azure.scale.test.huge.filesize=128M
The algorithm is
Only a few properties can be set this way; more will be added.
| Property | Meaninging |
|---|---|
| fs.azure.scale.test.huge.filesize | Size for huge file uploads |
| fs.azure.scale.test.huge.huge.partitionsize | Size for partitions in huge file uploads |
The file and partition sizes are numeric values with a k/m/g/t/p suffix depending on the desired size. For example: 128M, 128m, 2G, 2G, 4T or even 1P.
Some scale tests perform multiple operations (such as creating many directories).
The exact number of operations to perform is configurable in the option scale.test.operation.count
<property> <name>scale.test.operation.count</name> <value>10</value> </property>
Larger values generate more load, and are recommended when testing locally, or in batch runs.
Smaller values results in faster test runs, especially when the object store is a long way away.
Operations which work on directories have a separate option: this controls the width and depth of tests creating recursive directories. Larger values create exponentially more directories, with consequent performance impact.
<property> <name>scale.test.directory.count</name> <value>2</value> </property>
DistCp tests targeting Azure support a configurable file size. The default is 10 MB, but the configuration value is expressed in KB so that it can be tuned smaller to achieve faster test runs.
<property> <name>scale.test.distcp.file.size.kb</name> <value>10240</value> </property>
Azure-specific scale test properties are
The Huge File tests validate Azure storages’s ability to handle large files —the property fs.azure.scale.test.huge.filesize declares the file size to use.
<property> <name>fs.azure.scale.test.huge.filesize</name> <value>200M</value> </property>
Tests at this scale are slow: they are best executed from hosts running in the cloud infrastructure where the storage endpoint is based.
A selection of tests can run against the Azure Storage Emulator which is a high-fidelity emulation of live Azure Storage. The emulator is sufficient for high-confidence testing. The emulator is a Windows executable that runs on a local machine.
To use the emulator, install Azure SDK 2.3 and start the storage emulator. Then, edit src/test/resources/azure-test.xml and add the following property:
<property> <name>fs.azure.test.emulator</name> <value>true</value> </property>
There is a known issue when running tests with the emulator. You may see the following failure message:
com.microsoft.windowsazure.storage.StorageException: The value for one of the HTTP headers is not in the correct format.
To resolve this, restart the Azure Emulator. Ensure it is v3.2 or later.
Logging at debug level is the standard way to provide more diagnostics output; after setting this rerun the tests
log4j.logger.org.apache.hadoop.fs.azure=DEBUG
New tests are always welcome. Bear in mind that we need to keep costs and test time down, which is done by
No duplication: if an operation is tested elsewhere, don’t repeat it. This applies as much for metadata operations as it does for bulk IO. If a new test case is added which completely obsoletes an existing test, it is OK to cut the previous one —after showing that coverage is not worsened.
Efficient: prefer the getFileStatus() and examining the results, rather than call to exists(), isFile(), etc.
Fail with useful information: provide as much diagnostics as possible on a failure. Using org.apache.hadoop.fs.contract.ContractTestUtils to make assertions about the state of a filesystem helps here.
Isolating Scale tests. Any test doing large amounts of IO MUST extend the class AbstractAzureScaleTest, so only running if scale is defined on a build, supporting test timeouts configurable by the user. Scale tests should also support configurability as to the actual size of objects/number of operations, so that behavior at different scale can be verified.
Designed for parallel execution. A key need here is for each test suite to work on isolated parts of the filesystem. Subclasses of AbstractWasbTestBase SHOULD use the path(), methodpath() and blobpath() methods, to build isolated paths. Tests MUST NOT assume that they have exclusive access to a bucket.
Extending existing tests where appropriate. This recommendation goes against normal testing best practise of “test one thing per method”. Because it is so slow to create directory trees or upload large files, we do not have that luxury. All the tests against real endpoints are integration tests where sharing test setup and teardown saves time and money.
A standard way to do this is to extend existing tests with some extra predicates, rather than write new tests. When doing this, make sure that the new predicates fail with meaningful diagnostics, so any new problems can be easily debugged from test logs.
This is what we expect from new tests; they’re an extension of the normal Hadoop requirements, based on the need to work with remote servers whose use requires the presence of secret credentials, where tests may be slow, and where finding out why something failed from nothing but the test output is critical.
There are a set of base classes which should be extended for Azure tests and integration tests.
This extends the junit Assert class with thread names and timeouts, the default timeout being set in AzureTestConstants.AZURE_TEST_TIMEOUT to ten minutes. The thread names are set to aid analyzing the stack trace of a test: a jstack call can be used to
The base class for tests which use AzureBlobStorageTestAccount to create mock or live Azure clients; in test teardown it tries to clean up store state.
This class requires subclasses to implement createTestAccount() to create a mock or real test account.
The configuration used to create a test account should be that from createConfiguration(); this can be extended in subclasses to tune the settings.
This extends AbstractWasbTestBase for scale tests; those test which only run when -Dscale is used to select the “scale” profile. These tests have a timeout of 30 minutes, so as to support slow test runs.
Having shared base classes help reduces future maintenance. Please use them.
Don’t ever log credentials. The credential tests go out of their way to not provide meaningful logs or assertion messages precisely to avoid this.
This means efficient in test setup/teardown, and, ideally, making use of existing public datasets to save setup time and tester cost.
The reference example is ITestAzureHugeFiles:. This marks the test suite as @FixMethodOrder(MethodSorters.NAME_ASCENDING) then orders the test cases such that each test case expects the previous test to have completed (here: uploaded a file, renamed a file, …). This provides for independent tests in the reports, yet still permits an ordered sequence of operations. Do note the use of Assume.assume() to detect when the preconditions for a single test case are not met, hence, the tests become skipped, rather than fail with a trace which is really a false alarm.
As well as making file size and operation counts scaleable, this includes making test timeouts adequate. The Scale tests make this configurable; it’s hard coded to ten minutes in AbstractAzureIntegrationTest(); subclasses can change this by overriding getTestTimeoutMillis().
Equally importantly: support proxies, as some testers need them.
The ContractTestUtils class contains a whole set of assertions for making statements about the expected state of a filesystem, e.g. assertPathExists(FS, path), assertPathDoesNotExists(FS, path), and others. These do their best to provide meaningful diagnostics on failures (e.g. directory listings, file status, …), so help make failures easier to understand.
At the very least, do not use assertTrue() or assertFalse() without including error messages.
Although the auth-keys.xml file is marged as ignored in git and subversion, it is still in your source tree, and there’s always that risk that it may creep out.
You can avoid this by keeping your keys outside the source tree and using an absolute XInclude reference to it.
<configuration>
<include xmlns="http://www.w3.org/2001/XInclude"
href="file:///users/qe/.auth-keys.xml" />
</configuration>
The Azure tests create containers with the prefix "wasbtests-" and delete them after the test runs. If a test run is interrupted, these containers may not get deleted. There is a special test case which can be manually invoked to list and delete these, CleanupTestContainers
mvn test -Dtest=CleanupTestContainers
This will delete the containers; the output log of the test run will provide the details and summary of the operation.
Azure Data Lake Storage Gen 2 (ADLS Gen 2) is a set of capabilities dedicated to big data analytics, built on top of Azure Blob Storage. The ABFS and ABFSS schemes target the ADLS Gen 2 REST API, and the WASB and WASBS schemes target the Azure Blob Storage REST API. ADLS Gen 2 offers better performance and scalability. ADLS Gen 2 also offers authentication and authorization compatible with the Hadoop Distributed File System permissions model when hierarchical namespace is enabled for the storage account. Furthermore, the metadata and data produced by ADLS Gen 2 REST API can be consumed by Blob REST API, and vice versa.
In order to test ABFS, please add the following configuration to your src/test/resources/azure-auth-keys.xml file. Note that the ABFS tests include compatibility tests which require WASB credentials, in addition to the ABFS credentials.
<?xml version="1.0" encoding="UTF-8"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<configuration xmlns:xi="http://www.w3.org/2001/XInclude">
<property>
<name>fs.azure.abfs.account.name</name>
<value>{ACCOUNT_NAME}.dfs.core.windows.net</value>
</property>
<property>
<name>fs.azure.account.key.{ACCOUNT_NAME}.dfs.core.windows.net</name>
<value>{ACCOUNT_ACCESS_KEY}</value>
</property>
<property>
<name>fs.azure.wasb.account.name</name>
<value>{ACCOUNT_NAME}.blob.core.windows.net</value>
</property>
<property>
<name>fs.azure.account.key.{ACCOUNT_NAME}.blob.core.windows.net</name>
<value>{ACCOUNT_ACCESS_KEY}</value>
</property>
<property>
<name>fs.contract.test.fs.abfs</name>
<value>abfs://{CONTAINER_NAME}@{ACCOUNT_NAME}.dfs.core.windows.net</value>
<description>A file system URI to be used by the contract tests.</description>
</property>
<property>
<name>fs.contract.test.fs.wasb</name>
<value>wasb://{CONTAINER_NAME}@{ACCOUNT_NAME}.blob.core.windows.net</value>
<description>A file system URI to be used by the contract tests.</description>
</property>
</configuration>
To run OAuth and ACL test cases you must use a storage account with the hierarchical namespace enabled, and set the following configuration settings:
<!--=========================== AUTHENTICATION OPTIONS ===================-->
<!--ATTENTION:
TO RUN ABFS & WASB COMPATIBILITY TESTS, YOU MUST SET AUTH TYPE AS SharedKey.
OAUTH IS INTRODUCED TO ABFS ONLY.-->
<property>
<name>fs.azure.account.auth.type.{YOUR_ABFS_ACCOUNT_NAME}</name>
<value>{AUTH TYPE}</value>
<description>The authorization type can be SharedKey, OAuth, or Custom. The
default is SharedKey.</description>
</property>
<!--============================= FOR OAUTH ===========================-->
<!--IF AUTH TYPE IS SET AS OAUTH, FOLLOW THE STEPS BELOW-->
<!--NOTICE: AAD client and tenant related properties can be obtained through Azure Portal-->
<!--1. UNCOMMENT BELOW AND CHOOSE YOUR OAUTH PROVIDER TYPE -->
<!--
<property>
<name>fs.azure.account.oauth.provider.type.{ABFS_ACCOUNT_NAME}</name>
<value>org.apache.hadoop.fs.azurebfs.oauth2.{Token Provider Class name}</value>
<description>The full name of token provider class name.</description>
</property>
-->
<!--2. UNCOMMENT BELOW AND SET CREDENTIALS ACCORDING TO THE PROVIDER TYPE-->
<!--2.1. If "ClientCredsTokenProvider" is set as key provider, uncomment below and
set auth endpoint, client id and secret below-->
<!--
<property>
<name>fs.azure.account.oauth2.client.endpoint.{ABFS_ACCOUNT_NAME}</name>
<value>https://login.microsoftonline.com/{TENANTID}/oauth2/token</value>
<description>Token end point, this can be found through Azure portal</description>
</property>
<property>
<name>fs.azure.account.oauth2.client.id.{ABFS_ACCOUNT_NAME}</name>
<value>{client id}</value>
<description>AAD client id.</description>
</property>
<property>
<name>fs.azure.account.oauth2.client.secret.{ABFS_ACCOUNT_NAME}</name>
<value>{client secret}</value>
</property>
-->
<!--2.2. If "UserPasswordTokenProvider" is set as key provider, uncomment below and
set auth endpoint, use name and password-->
<!--
<property>
<name>fs.azure.account.oauth2.client.endpoint.{ABFS_ACCOUNT_NAME}</name>
<value>https://login.microsoftonline.com/{TENANTID}/oauth2/token</value>
<description>Token end point, this can be found through Azure portal</description>
</property>
<property>
<name>fs.azure.account.oauth2.user.name.{ABFS_ACCOUNT_NAME}</name>
<value>{user name}</value>
</property>
<property>
<name>fs.azure.account.oauth2.user.password.{ABFS_ACCOUNT_NAME}</name>
<value>{user password}</value>
</property>
-->
<!--2.3. If "MsiTokenProvider" is set as key provider, uncomment below and
set tenantGuid and client id.-->
<!--
<property>
<name>fs.azure.account.oauth2.msi.tenant.{ABFS_ACCOUNT_NAME}</name>
<value>{tenantGuid}</value>
<description>msi tenantGuid.</description>
</property>
<property>
<name>fs.azure.account.oauth2.client.id.{ABFS_ACCOUNT_NAME}</name>
<value>{client id}</value>
<description>AAD client id.</description>
</property>
-->
<!--2.4. If "RefreshTokenBasedTokenProvider" is set as key provider, uncomment below and
set refresh token and client id.-->
<!--
<property>
<name>fs.azure.account.oauth2.refresh.token.{ABFS_ACCOUNT_NAME}</name>
<value>{refresh token}</value>
<description>refresh token.</description>
</property>
<property>
<name>fs.azure.account.oauth2.client.id.{ABFS_ACCOUNT_NAME}</name>
<value>{client id}</value>
<description>AAD client id.</description>
</property>
-->
<!--
<property>
<name>fs.azure.identity.transformer.enable.short.name</name>
<value>true/false</value>
<description>
User principal names (UPNs) have the format “{alias}@{domain}”.
If true, only {alias} is included when a UPN would otherwise appear in the output
of APIs like getFileStatus, getOwner, getAclStatus, etc, default is false.
</description>
</property>
<property>
<name>fs.azure.identity.transformer.domain.name</name>
<value>domain name of the user's upn</value>
<description>
If the domain name is specified and “fs.azure.identity.transformer.enable.short.name”
is true, then the {alias} part of a UPN can be specified as input to APIs like setOwner,
setAcl, modifyAclEntries, or removeAclEntries, and it will be transformed to a UPN by appending @ and the domain specified by
this configuration property.
</description>
</property>
<property>
<name>fs.azure.identity.transformer.service.principal.id</name>
<value>service principal object id</value>
<description>
An Azure Active Directory object ID (oid) used as the replacement for names contained
in the list specified by “fs.azure.identity.transformer.service.principal.substitution.list”.
Notice that instead of setting oid, you can also set $superuser here.
</description>
</property>
<property>
<name>fs.azure.identity.transformer.skip.superuser.replacement</name>
<value>true/false</value>
<description>
If false, “$superuser” is replaced with the current user when it appears as the owner
or owning group of a file or directory. The default is false.
</description>
</property>
<property>
<name>fs.azure.identity.transformer.service.principal.substitution.list</name>
<value>mapred,hdfs,yarn,hive,tez</value>
<description>
A comma separated list of names to be replaced with the service principal ID specified by
“fs.azure.identity.transformer.service.principal.id”. This substitution occurs
when setOwner, setAcl, modifyAclEntries, or removeAclEntries are invoked with identities
contained in the substitution list. Notice that when in non-secure cluster, asterisk symbol *
can be used to match all user/group.
</description>
</property>
-->
If running tests against an endpoint that uses the URL format http[s]://[ip]:[port]/[account]/[filesystem] instead of http[s]://[account][domain-suffix]/[filesystem], please use the following:
<property>
<name>fs.azure.abfs.endpoint</name>
<value>{IP}:{PORT}</value>
</property>