feat: use concrete classes for hbase.client.registry.impl (#4043) (#4188

)

Previously we used runtime bytecode generation which prevented endusers from using hbase-site.xml to target bigtable in hbase 2 applications. This PR will create a concrete implementation for both AsyncRegistry for (hbase < 2.3) and ConnectionRegistry (for hbase >= 2.3). 
To accomplish this, we use a local copy of the AsyncRegistry interface from HBase 2.2 that we use as a compile time target and then strip it from the resulting jar.

After this change end users should be able to use async apis in hbase 2 using hbase-site.xml. Examples:



```xml
<configuration>
    <property>
        <name>hbase.client.connection.impl</name>
        <value>com.google.cloud.bigtable.hbase2_x.BigtableConnection</value>
    </property>
    <property>
        <name>hbase.client.async.connection.impl</name>
        <value>org.apache.hadoop.hbase.client.BigtableAsyncConnection</value>
    </property>
    <property>
        <name>hbase.client.registry.impl</name>
        <!-- When using hbase >= 2.3, use this -->
        <value>org.apache.hadoop.hbase.client.BigtableConnectionRegistry</value>
        <!-- When using hbase < 2.3, use this -->
        <!-- <value>org.apache.hadoop.hbase.client.BigtableAsyncRegistry</value> -->
    </property>
    <property>
        <name>google.bigtable.project.id</name>
        <value>my-project</value>
    </property>
    <property>
        <name>google.bigtable.instance.id</name>
        <value>my-instance</value>
    </property>
    <property>
        <name>google.bigtable.app_profile.id</name>
        <value>my-app-profile</value>
    </property>
</configuration>
```

```java
AsyncConnection asyncConnection = ConnectionFactory.createAsyncConnection().get();
```

bigtable-client-core-parent/bigtable-hbase/pom.xml

@@ -176,12 +176,6 @@ limitations under the License.
       <scope>runtime</scope>
     </dependency>
 
-    <dependency>
-      <groupId>net.bytebuddy</groupId>
-      <artifactId>byte-buddy</artifactId>
-      <version>${byte.buddy.version}</version>
-    </dependency>
-
     <!-- Test -->
     <dependency>
       <groupId>com.google.cloud.bigtable</groupId>

bigtable-client-core-parent/bigtable-hbase/src/main/java/com/google/cloud/bigtable/hbase/BigtableConfiguration.java

@@ -20,7 +20,6 @@
 import com.google.auth.Credentials;
 import com.google.common.base.Preconditions;
 import org.apache.hadoop.conf.Configuration;
-import org.apache.hadoop.hbase.client.BigtableAsyncRegistry;
 import org.apache.hadoop.hbase.client.Connection;
 
 /** This class provides a simplified mechanism of creating a programmatic Bigtable Connection. */
@@ -35,34 +34,20 @@ public class BigtableConfiguration {
   public static final String HBASE_CLIENT_ASYNC_CONNECTION_IMPL =
       "hbase.client.async.connection.impl";
 
+  // The value of this field can implement 2 different interfaces depending on the HBase version
+  // For 2.0 - 2.2 the value must implement AsyncRegistry
+  // For 2.3 onwards the value must implement ConnectionRegistry
+  // bigable-hbase-2x contains implementations for both, this helper will use classpath probing to
+  // guess the correct impl.
   /** For internal use only - public for technical reasons. */
   @InternalApi("For internal usage only")
   public static final String HBASE_CLIENT_ASYNC_REGISTRY_IMPL = "hbase.client.registry.impl";
 
-  /** For internal use only - public for technical reasons. */
-  @InternalApi("For internal usage only")
-  public static final String BIGTABLE_HBASE_CLIENT_ASYNC_CONNECTION_CLASS =
-      "org.apache.hadoop.hbase.client.BigtableAsyncConnection";
-
-  /** For internal use only - public for technical reasons. */
-  @InternalApi("For internal usage only")
-  public static final String BIGTABLE_HBASE_CLIENT_ASYNC_REGISTRY_CLASS =
-      "org.apache.hadoop.hbase.client.BigtableAsyncRegistry";
-
-  private static final String HBASE_ASYNC_CONNECTION_CLASS =
-      "org.apache.hadoop.hbase.client.AsyncConnection";
-
   private static final String[] CONNECTION_CLASS_NAMES = {
     "com.google.cloud.bigtable.hbase1_x.BigtableConnection",
     "com.google.cloud.bigtable.hbase2_x.BigtableConnection",
   };
 
-  private static final String[] ASYNC_CONNECTION_CLASS_NAMES = {
-    BIGTABLE_HBASE_CLIENT_ASYNC_CONNECTION_CLASS,
-    BIGTABLE_HBASE_CLIENT_ASYNC_REGISTRY_CLASS,
-    HBASE_ASYNC_CONNECTION_CLASS
-  };
-
   private static final Class<? extends Connection> CONNECTION_CLASS = chooseConnectionClass();
 
   @SuppressWarnings("unchecked")
@@ -91,24 +76,57 @@ public static Class<? extends Connection> getConnectionClass() {
     return CONNECTION_CLASS;
   }
 
+  private static Class<?> getConnectionRegistryClass() {
+    try {
+      Class.forName("org.apache.hadoop.hbase.client.ConnectionRegistry");
+      return Class.forName("org.apache.hadoop.hbase.client.BigtableConnectionRegistry");
+    } catch (ClassNotFoundException e) {
+      // noop
+    }
+
+    try {
+      Class.forName("org.apache.hadoop.hbase.client.AsyncRegistry");
+      return Class.forName("org.apache.hadoop.hbase.client.BigtableAsyncRegistry");
+    } catch (ClassNotFoundException e) {
+      // noop
+    }
+
+    return null;
+  }
+
+  private static Class<?> getAsyncConnectionClass() {
+    try {
+      // Make sure HBase interface is present
+      Class.forName("org.apache.hadoop.hbase.client.AsyncConnection");
+    } catch (ClassNotFoundException e) {
+      return null;
+    }
+
+    try {
+      // then try loading our implementation
+      return Class.forName("org.apache.hadoop.hbase.client.BigtableAsyncConnection");
+    } catch (ClassNotFoundException e) {
+      return null;
+    }
+  }
+
   /**
    * Set up connection impl classes. If Bigtable and HBase async connection classes exist, set up
    * async connection impl class as well.
    */
   private static Configuration injectBigtableImpls(Configuration configuration) {
+    // HBase 1x & 2x sync connection
     configuration.set(HBASE_CLIENT_CONNECTION_IMPL, getConnectionClass().getCanonicalName());
-    try {
-      // Set up HBase async registry impl if async connection classes exist
-      for (String className : ASYNC_CONNECTION_CLASS_NAMES) {
-        Class.forName(className);
-      }
-      configuration.set(
-          HBASE_CLIENT_ASYNC_CONNECTION_IMPL, BIGTABLE_HBASE_CLIENT_ASYNC_CONNECTION_CLASS);
-      configuration.set(
-          HBASE_CLIENT_ASYNC_REGISTRY_IMPL, BigtableAsyncRegistry.getSubClass().getName());
-    } catch (ClassNotFoundException ignored) {
-      // Skip if any of the async connection class doesn't exist
+
+    // HBase 2x async
+    Class<?> asyncConnectionClass = getAsyncConnectionClass();
+    Class<?> connectionRegistryClass = getConnectionRegistryClass();
+
+    if (asyncConnectionClass != null && connectionRegistryClass != null) {
+      configuration.set(HBASE_CLIENT_ASYNC_CONNECTION_IMPL, asyncConnectionClass.getName());
+      configuration.set(HBASE_CLIENT_ASYNC_REGISTRY_IMPL, connectionRegistryClass.getName());
     }
+
     return configuration;
   }
 
@@ -179,25 +197,6 @@ public static Configuration withCredentials(Configuration conf, Credentials cred
     return new BigtableExtendedConfiguration(conf, credentials);
   }
 
-  /**
-   * Configuration for getting a org.apache.hadoop.hbase.client.AsyncConnection.
-   *
-   * @param conf a {@link org.apache.hadoop.conf.Configuration} object to configure.
-   * @return the modified {@link org.apache.hadoop.conf.Configuration} object.
-   * @deprecated For HBase 2.x async connections, please use {@link #configure(String, String)}
-   *     directly.
-   *     <pre>{@code
-   * Configure config = BigtableConfiguration.configure(projectId, instanceId);
-   * AsyncConnection asyncConnection = ConnectionFactory.createAsyncConnection(config).get();
-   * }</pre>
-   */
-  @Deprecated
-  public static Configuration asyncConfigure(Configuration conf) {
-    conf.set(HBASE_CLIENT_ASYNC_CONNECTION_IMPL, BIGTABLE_HBASE_CLIENT_ASYNC_CONNECTION_CLASS);
-    conf.set(HBASE_CLIENT_ASYNC_REGISTRY_IMPL, BigtableAsyncRegistry.getSubClass().getName());
-    return conf;
-  }
-
   /**
    * connect.
    *

bigtable-hbase-2.x-parent/bigtable-hbase-2.x-hadoop/pom.xml

@@ -251,6 +251,7 @@ limitations under the License.
                 <classEntry>org/apache/hadoop/hbase/client/AbstractBigtableAdmin</classEntry>
                 <classEntry>org/apache/hadoop/hbase/client/AbstractBigtableConnection</classEntry>
                 <classEntry>org/apache/hadoop/hbase/client/BigtableAsyncConnection</classEntry>
+                <classEntry>org/apache/hadoop/hbase/client/BigtableConnectionRegistry</classEntry>
                 <classEntry>org/apache/hadoop/hbase/client/BigtableAsyncRegistry</classEntry>
                 <classEntry>org/apache/hadoop/hbase/client/CommonConnection</classEntry>
               </allowedJarClassEntries>

bigtable-hbase-2.x-parent/bigtable-hbase-2.x-shaded/pom.xml

@@ -431,8 +431,10 @@ limitations under the License.
                 <classEntry>org/apache/hadoop/hbase/client/AbstractBigtableAdmin</classEntry>
                 <classEntry>org/apache/hadoop/hbase/client/AbstractBigtableConnection</classEntry>
                 <classEntry>org/apache/hadoop/hbase/client/BigtableAsyncConnection</classEntry>
+                <classEntry>org/apache/hadoop/hbase/client/BigtableConnectionRegistry</classEntry>
                 <classEntry>org/apache/hadoop/hbase/client/BigtableAsyncRegistry</classEntry>
                 <classEntry>org/apache/hadoop/hbase/client/CommonConnection</classEntry>
+                <classEntry>org/apache/hadoop/hbase/client/BigtableConnectionRegistry</classEntry>
               </allowedJarClassEntries>
             </configuration>
           </execution>

bigtable-hbase-2.x-parent/bigtable-hbase-2.x/pom.xml

@@ -194,6 +194,22 @@ limitations under the License.
     </dependency>
   </dependencies>
   <build>
+    <pluginManagement>
+      <plugins>
+        <!-- AsyncRegistry was an interface that existed in hbase 2.0-2.2. It's no longer shipped. To ensure
+        compatibility we vendored it in this project to compile against, but do not want to ship it our binary -->
+        <plugin>
+          <groupId>org.apache.maven.plugins</groupId>
+          <artifactId>maven-jar-plugin</artifactId>
+          <configuration>
+            <excludes>
+              <exclude>org/apache/hadoop/hbase/client/AsyncRegistry.class</exclude>
+            </excludes>
+          </configuration>
+        </plugin>
+      </plugins>
+    </pluginManagement>
+
     <plugins>
       <plugin>
         <groupId>org.apache.maven.plugins</groupId>
@@ -205,6 +221,9 @@ limitations under the License.
                   See bigtable-hbase-1.x pom.xml for more details-->
             <ignoredNonTestScopedDependency>*</ignoredNonTestScopedDependency>
           </ignoredNonTestScopedDependencies>
+          <ignoredDependencies>
+            <dependency>*</dependency>
+          </ignoredDependencies>
         </configuration>
       </plugin>
       <plugin>
@@ -215,6 +234,25 @@ limitations under the License.
           <skip>true</skip>
         </configuration>
       </plugin>
+
+      <plugin>
+        <groupId>org.codehaus.mojo</groupId>
+        <artifactId>build-helper-maven-plugin</artifactId>
+        <executions>
+          <execution>
+            <id>add-source</id>
+            <phase>generate-sources</phase>
+            <goals>
+              <goal>add-source</goal>
+            </goals>
+            <configuration>
+              <sources>
+                <source>../../third_party/third_party_hbase_client_pre23/src/main/java</source>
+              </sources>
+            </configuration>
+          </execution>
+        </executions>
+      </plugin>
     </plugins>
   </build>
 </project>