CHANGES.md

Change History

1.2.2 (2018-05-16)

• Bugfix: writer thread was not a daemon, would keep application from shutting down unless explicitly terminated. (#38)

1.2.1 (2018-01-06)

• Use reflection to create AWS service clients from default factory methods. This will be the default behavior for most SDK versions (unless explicit clientFactory or endpoint configured). (#30
• Added the clientFactory property, which instructs appenders to call a static factory method to create their AWS service client. (#28
• Allow explicit endpoint configuration. This is intended to support clients using older AWS SDKs that don't want to direct output to the us-east-1 region. (#30
• When creating a service client via constructor, attempt to set the region from the AWS_REGION environment variable. This is an alternative to specifying endpoint in your logger config, but in my opinion isn't very useful because the list of available regions is dependent on your SDK version. For example, SDK 1.11.0 doesn't know about the us-east-2 region. (#30
• Enable per-record random partition keys for KinesisAppender. This can make better use of a multi-shard stream for applications with high logging output. (#24

1.2.0 (2017-12-30)

• Added SNS as a destination.
• CloudWatchAppender and KinesisAppender now refuse to initialize if given an invalid stream/group name. In earlier revisions they would try to strip out invalid characters, but this meant that you could send messages to the wrong place. Better to fail fast.
• If unable to initialize, set internal queue to discard all messages. This avoids unnecessary memory consumption for messages that will never be sent.
• Allowed discard threshold and action to be reconfigured at runtime. This was added to support handling initialization errors, but may be useful for some applications.

1.1.3 (2017-12-12)

• Bugfix: CloudWatchAppender now handles multi-part results from DescribeLogGroups and DescribeLogStreams. This was unlikely to happen in the real world, but could affect sites with large numbers of similarly-named groups or streams.
• Bugfix: KinesisAppender is now more resilient to rate limiting during initialization. This was primarily an issue with the integration tests, which create and delete streams at a high rate. In normal operations it could happen if a large number of applications started logging at the same time (such as a fleet of instances starting).
• AbstractLogWriter now tracks whether an error happened during initialization. This is currently used for testing, but will be exposed via JMX at some future point.

1.1.2 (2017-10-16)

• Added JsonLayout.
• Added an example application that sends output to all appenders.

1.1.1 (2017-10-10)

• Added configurable message discard, to avoid out-of-memory errors when connectivity to AWS is spotty. (#15)

1.1.0 (2017-09-29)

• Added Kinesis as a destination.
• Improved implementation of CloudWatchAppender, including integration tests.

1.0.1 (2017-09-02)

• Bugfix: CloudWatch SDK dependency not marked as provided; might cause conflict with a project's own dependency versions.
• Bugfix: Default configuration was incorrect: log stream name was "{startTimestamp}", should be "{startupTimestamp}"
• Bugix: more resilient to rate limiting of logstream/loggroup creation.

1.0.0 (2017-08-12)

• Initial release: CloudWatch Logs as a destination.