The Apache Accumulo community has announced its 1.7.0 release. As community’s first major release of 2015, the release represents the culmination of a year of effort from many Accumulo committers and contributors. Apart from many notable changes enumerated below, Accumulo is now well integrated with Apache Ambari.
In this release, 43 different individuals fixed 691 JIRA issues, and we thank everyone who helped in any way to make this Apache Accumulo 1.7.0 a reality.
Accumulo is a strongly consistent and durable store. One drawback to this implementation is that this approach makes it very difficult to obtain a separate complete copy of the data in Accumulo without the client writing it to a second instance. ACCUMULO-378 first appears in Apache with 1.7.0 as a means for users to configure one Accumulo instance to automatically copy data to another Accumulo instance. This copy is performed in an eventually consistent manner, ideal for standby or disaster-recovery configurations.
Kerberos is the means of choice to enable strong authentication across Hadoop and related applications. While Accumulo could previously operate on secured Hadoop clusters, it lacked the ability for Accumulo clients to interact with the system using their Kerberos credentials. ACCUMULO-2815 adds support for users to authenticate with their Kerberos credentials via the Java public API or MapReduce Accumulo APIs.
Accumulo’s previous metrics implementation was deprecated in favor of using Hadoop Metrics2. Metrics2 defines a system of sources and sinks for handling metrics: sources generate metrics while sinks consume and persist them. Most notable are the provided sink implementations, which include Ganglia, Graphite and the new Ambari Metrics System (AMBARI-5707 and AMBARI-10144).
Despite Accumulo long-having Dapper-style distributed tracing support, the growth of HTrace, especially in HDFS, meant that using a custom implementation would prevent Accumulo from reaping the benefits of great HDFS metrics that enrich the existing server instrumentations. Adopting HTrace benefits Accumulo directly with a wealth of new metrics data in addition to contributing back to HDFS and HTrace to ensure a better and more informative tracing system for all users.
Apache Ambari is a provisioning, managing and monitoring application for Apache Hadoop clusters. AMBARI-1930, slated for Apache Ambari 2.1.0, provides support for the installation, configuration and monitoring of Apache Accumulo 1.7.0. Accumulo users will finally be able to reap the benefits of a centralized service dedicated to the management of Apache Accumulo and the rest of their Hadoop cluster via Ambari.
Through other features delivered in Apache Accumulo 1.7.0, users can alter Accumulo configurations on the fly, start and stop Accumulo processes, or view important Accumulo metrics, all through a web browser.
As always, there is much in store for the next version of Apache Accumulo. Some new changes we can expect in the near future are:
For more information on Apache Accumulo 1.7.0, the following resources are available: