bigtable-client-0.9.5.1

• Added a Cloud Dataflow-based export tool, and fixed the Cloud Dataflow-based import tool. See the GitHub README for details.
• Improved the performance of scans.
• Updated gRPC to version 1.0.3.

bigtable-client-0.9.4

This version includes the following changes:

• Fixed an issue that resulted in increased CPU usage over time by reverting gRPC to 0.14.1 and Netty to 4.1.1.Final. This issue was introduced in the 0.9.3 release.
• When there is a temporary authentication issue, the client now refreshes credentials in addition to refreshing the OAuth 2.0 token.
• Fixed an issue that caused BufferedMutator#flush() to never complete.
• Fixed a regression in performance in BufferedMutator that reduced throughput.
• In the Cloud Dataflow connector, fixed an edge case in which dynamic splitting caused exceptions.

bigtable-client-0.9.3

This version includes the following changes:

• Fixed performance issues when reading large rows. These issues were introduced in the 0.9.2 release.
  Added utilities to resize clusters programmatically:

  • To set a specific number of nodes:

    try (BigtableClusterUtilities clusterUtilities = 
     BigtableClusterUtilities.forInstance([PROJECT_ID], [INSTANCE_ID])) {
      clusterUtilities.setClusterSize([POSITIVE_INT]);
    }

  • To increment or decrement the number of nodes:

    try (BigtableClusterUtilities clusterUtilities =
      BigtableClusterUtilities.forInstance([PROJECT_ID], [INSTANCE_ID])) {
      clusterUtilities.incrementClusterSize([NEGATIVE_OR_POSITIVE_INT]);
    }

• Re-added dynamic rebalancing to the Cloud Dataflow connector. Dynamic rebalancing allows Cloud Dataflow to split shards that are performing slowly into smaller shards, allowing the job to finish faster.
• Added infinite retries for temporary authentication issues.
• Removed the requirement for credentials when testing against the Cloud Bigtable emulator.
• Added a new timer (google-cloud-bigtable.grpc.method.ReadRows.firstResponse.latency) for the latency of receiving the first response to a scan request. See "Reporting Cloud Bigtable Metrics" for details about using client-side metrics.

bigtable-client-0.9.2

Do not use this version. Use 0.9.3 instead. This version has a performance degradation when reading large rows.

• Integrated with Dropwizard Metrics to add client-side metrics. See "Reporting Cloud Bigtable Metrics" for details.
• Added improved support for the Cloud Bigtable emulator. See "Using the Cloud Bigtable Emulator" for details.
• Removed Cloud Dataflow dynamic rebalancing, which caused some rows to be read twice.
• Upgraded gRPC to 1.0.0, improving performance and stability.
• Removed protobuf files and Java artifacts for the v1 API.

bigtable-client-0.9.1

• You must now use version 1.1.33.Fork19 of the netty-tcnative-boringssl-static library.
• You can now use a single netty-tcnative-boringssl-static JAR file for all supported platforms (Linux, OS X, and Windows). See "Setting up encryption" for instructions on how to set up your pom.xml file.
• Added two new options, which you can set in your hbase-site.xml file or programmatically:
  • google.bigtable.rpc.use.timeouts: Determines whether RPCs will time out after a specified number of milliseconds. Set to false (default) or true.
  • google.bigtable.rpc.timeout.ms: The timeout value, in milliseconds. Defaults to 60000 (1 minute) for consistency with HBase. For small gets and puts, 2000 is a reasonable timeout value. For other workloads, you will need to experiment to find an appropriate timeout value.
  • Fixed an issue that caused IPv6 addresses to be used in environments that do not support IPv6, such as Docker containers that are configured to use only IPv4.
• Calls to getTable() are now five to ten times faster. (changed from microseconds to hundreds of nanoseconds)
• Added a BigtableInstanceClient, accessible from BigtableSession, that you can use for administering your instance and cluster.

bigtable-client-0.9.0

• Updated implementation to use the new Cloud Bigtable instance API and removed the cluster API implementation. Using this new client will require changing the configuration to work with the new API.
• Updated Cloud Dataflow version to 1.6
• Added dynamic rebalancing integration in the Cloud Dataflow's support for reading from Cloud Bigtable. Dynamic rebalancing allows Cloud Dataflow to split shards that are performing slower than other shards into smaller shards so that the job finishes faster, as described in this blog post.
• Added support for FuzzyRowFilter

version 0.3.0

• The method BufferedMutator#mutate() now uses bulk mutations, which improve throughput.
• You can now perform bulk mutations without using the HBase API by calling the method BigtableSession#createBulkMutation().
• Bulk mutations have been made more robust by retrying failed individual mutations in bulk.
• You can now perform bulk reads without using the HBase API by calling the method BigtableSession#createBulkRead().
• You can now delete row ranges in bulk by calling the method BigtableTableAdminGrpcClient#bulkDeleteRows() or using the HBase API's truncateTable() methods. In addition, the HBase API's truncateTable() methods no longer drop and recreate the table.
• The method AbstractCloudBigtableTableFn#retrowException() has been renamed to AbstractCloudBigtableTableFn#rethrowException(). If you have created a subclass of AbstractCloudBigtableTableFn that uses this method, you must rename the method in your subclass.
• You can now call the method org.apache.hadoop.hbase.client.Admin#modifyColumn() to alter an existing column. In addition, the HBase shell's alter command now allows you to alter an existing column.
• The Google authentication library was updated to fix issues with long-running jobs in App Engine's flexible environment.

version 0.2.4

• Retry logic is improved for scans, mutations (puts), and gets. Note: All HBase Puts are now automatically retried even if no timestamp is set.
• Throughput is significantly increased for asynchronous (BufferedMutator) operations.
• Latency is reduced for synchronous operations.
• Resolved several dependency contentions when using the client with the gRPC library.
• Cloud Bigtable operations, such as the HBase shell quickstart, can now be used in Google Cloud Shell.

Cloud Dataflow connector

• You can now use the Cloud Dataflow connector with Cloud Dataflow 1.5. In addition, the connector uses larger batch sizes when you use Cloud Dataflow 1.5.
• Added a utility class, AbstractCloudBigtableDoFn, to the Cloud Dataflow connector. You can use this utility class to get a BigtableConnection for doing additional Cloud Bigtable work. For example, you can use this class to create a DoFn that performs additional Gets while doing a table scan.
• Added a new static method, com.google.cloud.bigtable.dataflow.CloudBigtableIO.readBulk(), that can read an array of elements at a time. You can use the array of elements to perform more efficient bulk processing than possible for a single element.
• You can now use Cloud Dataflow to import Hadoop sequence files generated by HBase.
• The Cloud Dataflow connector can now use Cloud Bigtable Sources across more workers. A Source will now support up to 4,000 splits.
• The Cloud Dataflow connector no longer performs validation lookups, which were generating many unnecessary requests.

bigtable-client-0.2.3

Do not use this version. Use the latest instead. This version does not work in environments that have problems using IPv6, such as Docker and Dataflow.

version 0.2.2

• The hbase-client artifact is no longer a required dependency and should be removed from your pom.xml file.
• You can now use OpenSSL for encryption instead of adding the ALPN library to your boot classpath. Using OpenSSL typically results in a 5% to 20% performance improvement for encryption. See "Using OpenSSL encryption" for details.
• There is now a helper class that simplifies the process of creating a connection. See "Including connection settings in your code" for details.
• The RowFilter filter is now supported. With this filter, you can use either the BinaryComparator comparator or the RegexStringComparator comparator with no flags and the EQUAL operator.
• Fixed an issue that could cause deletes to hang forever during scans.
• When there are duplicate rows that contain the same row keys, column families, column qualifiers, and timestamps, the client now de-duplicates the rows.
• You can now use the HBase shell's alter command to delete column families.
• You can now use the HBase shell's truncate command, which disables, drops, and re-creates a table.