blob: 852bb0d849f4026dcd7a87ae0a6aaa3a78855e74 [file] [log] [blame]
* [Admin Dashboard] Added a web based GUI for exploring running and finished jobs in a running Gobblin daemon (thanks Eric Ogren).
* [Admin Dashboard] Added a CLI for finding jobs in the job history store and seeing their run details (thanks Eric Ogren).
* [Configuration Management] WIP: Configuration management library. Will enable Gobblin to be dataset aware, ie. to dynamically load and apply different configurations to each dataset in a single Gobblin job.
** APIs: APIs for configuration stores and configuration client.
** Configuration Library: loads low level configurations from a configuration store, resolves configuration dependencies / imports, and performs value interpolation.
* [Distcp] Allow using *.ready files as markers for files that should be copied, and deletion of *.ready files once the file has been copied.
* [Distcp] Added file filters to recursive copyable dataset for distcp. Allows to only copy files satisfying a filter under a base directory.
* [Distcp] Copied files that fail to be published are persisted for future runs. Future runs can recover the already copied file instead of re-doing the byte transfer.
* [JDBC] Can use password encryption for JDBC sources.
* [YARN] Added email notifications on YARN application shutdown.
* [YARN] Added event notifications on YARN container status changes.
* [Metrics] Added metric filters based on name and type of the metrics.
* [Dataset Management] POC embedded sql for config-driven retention management.
* [Exactly Once] POC for Gobblin managed exactly once semantics on publisher.
* **Core** File based source includes previously failed WorkUnits event if there are no new files in the source (thanks Joel Baranick).
* **Core** Ensure that output file list does not contain duplicates due to task retries (thanks Joel Baranick).
* **Core** Fix NPE in CliOptions.
* **Core/YARN** Limit Props -> Typesafe Config conversion to a few keys to prevent overwriting of certain properties.
* **Utility** Fixed writer mkdirs for S3.
* **Metrics** Made Scheduled Reporter threads into daemon threads to prevent hanging application.
* **Metrics** Fixed enqueuing of events on event reporters that was causing job failure if event frequency was too high.
* **Build** Fix POM dependencies on gobblin-rest-api.
* **Build** Added conjars and cloudera repository to all projects (fixes builds for certain users).
* **Build** Fix the distribution tarball creation (thanks Joel Baranick).
* **Build** Added option to exclude Hadoop and Hive jars from distribution tarball.
* **Build** Removed from runtime resources.
* **Compaction** Fixed main class in compaction manifest file (thanks Lorand Bendig).
* **JDBC** Correctly close JDBC connections.
* [Build] Add support for publishing libraries to maven local (thanks Joel Baranick).
* [Build] In preparation to Gradle 2 migration, added ext. prefix to custom gradle properties.
* [Build] Can generate project dependencies graph in dot format.
* [Metrics] Migrated Kafka reporter and Output stream reporter to Root Metrics Reporter managed reporting.
* [Metrics] The last metric emission in the application has a "final" tag for easier Hive identification.
* [Metrics] Metrics for Gobblin on YARN include cluster tags.
* [Hive] Upgraded Hive to version 1.0.1.
* [Distcp] Add file size to distcp success notifications.
* [Distcp] Each work unit in distcp contains exactly one Copyable File.
* [Distcp] Copy source can set upstream timestamps for SLA events emitted on publish time.
* [Scheduling] Added Gobblin Oozie config files.
* [Documentation] Improved javadocs.
- **Build/release** Adding build instrumentation for generation of rest-api-* artifacts
- **Build/release** Various fixes to decrease reliance of unit tests on timing.
- **Core** Add stability annotations for APIs. We plan on starting to annotate interfaces/classes to specify how likely the API is to change.
- **Runtime** Made it an option for the job scheduler to wait for running jobs to complete
- **Runtime** Fixing dangling MetricContext creation in ForkOperator
- kadaan, joel.baranick:
+ Added a fix for a hadoop issue ( which affects the s3a filesystem and results in duplicate files appearing in the results of ListStatus. In the process, extracted a base class for all FsHelper classes based on the hadoop filesystem.
* [Compaction] Added M/R compaction/de-duping for hourly data
* [Compaction] Added late data handling for hourly and daily M/R compaction:; added support for triggering M/R compaction if late data exceeds a threshold
* [I/O] Added support for using Hive SerDe's through HiveWritableHdfsDataWriter
* [I/O] Added the concept of data partitioning to writers:
* [Runtime] Added CliLocalJobLauncher for launching single jobs from the command line.
* [Converters] Added AvroSchemaFieldRemover that can remove specific fields from a (possibly recursive) Avro schema.
* [DQ] Added new row-level policies RecordTimestampLowerBoundPolicy and AvroRecordTimestampLowerBoundPolicy for checking if a record timestamp is too far in the past.
* [Kafka] Added schema registry API to KafkaAvroExtractor which enables supports for various Kafka schema registry implementations (e.g. Confluent's schema registry).
* [Build/Release] Added build instrumentation to publish artifacts to Maven Central
* [Retention management] Trash handles deletes of files already existing in trash correctly.
* [Kafka] Fixed an issue that may cause Kafka adapter to miss data if the fork fails.
* [Runtime] Added metrics for job executions
* [Metrics] Added a root metric context to keep track of GC of metrics and metric contexts and make sure those are properly reported
* [Compaction] Improve topic isolation in MRCompactor
* [Build/release] Java version compatibility raised to Java 7.
* [Runtime] Deprecated COMMIT_ON_PARTIAL_SUCCESS and added a new policy for successful extracts
* [Retention management] Async trash implementation for parallel deletions.
* [Metrics] Added tracking events emission when data gets published
* [Retention management] Added support for parallel execution to the dataset cleaner
* [Runtime] Update job execution info in the execution history store upon every task completion
Note: these are new features which are under active development and may be subject to significant changes.
* [gobblin-ce] Adding support for Gobblin Continuous Execution on Yarn
* [distcp-ng] Started work on bulk transfer (file copies) using Gobblin
* [distcp-ng] Added a light-weight Hadoop FileSystem implementation for file transfer from SFTP
* [gobblin-config] Added API for dataset driven
We would like to thank all our external contributors for helping improve Gobblin.
* kadaan, joel.baranick:
- Separate publisher filesystem from writer filesystem
- Support for generating Idea projects with the correct language level (Java 7)
- Fixed yarn conf path in
* mwol(Maurice Wolter)
- Implemented new class AvroCombineFileSplit which stores the avro schema for each split, determined by the corresponding input file.
* cheleb(NOUGUIER Olivier)
- Add support for maven install
* dvenkateshappa
- bugifx to
- Added an excluding column list , which can be used for salesforce configuration with huge list of columns.
* klyr (Julien Barbot)
- bugfix to
* gheo21
- Bumped kafka dependency to 2.11
* ahollenbach (Andrew Hollenbach)
- configuration improvements for standalone mode
* lbendig (Lorand Bendig)
- fixed a bug in DatasetState creation