monasca-persister/java
Witek Bedyk f7f3d2e9f8 Deprecate Java implementation
The change adds deprecation notice and updates Readme file.
Also include links to Storyboard, and notes on configuration.

Story: 2005628
Task: 30881
Change-Id: I0d7ae536d02db4bbcb155ad3a5d0b84761a31594
2019-09-17 10:29:34 +02:00
..
src Update all columns in metrics on an update to refresh TTL 2019-07-19 13:04:23 +02:00
pom.xml Disable ClassPathURLCheck for Java build 2018-11-06 14:43:01 +01:00
README.rst Deprecate Java implementation 2019-09-17 10:29:34 +02:00

monasca-persister

Warning

Java implementation of monasca-persister is deprecated as of Train release.

The Monasca Persister consumes metrics and alarm state transitions from the Apache Kafka message queue and stores them in the time series database.

Although the Persister isn't primarily a Web service it uses DropWizard, https://dropwizard.github.io/dropwizard/, which provides a nice Web application framework to expose an http endpoint that provides an interface through which metrics about the Persister can be queried as well as health status.

The basic design of the Persister is to have one Kafka consumer publish to a Disruptor, https://github.com/LMAX-Exchange/disruptor, that has output processors. The output processors use prepared batch statements to write to the Metrics and Alarms database.

The number of output processors/threads in the Persister can be specified to scale to more messages. To horizontally scale and provide fault-tolerance any number of Persisters can be started as consumers from the Message Queue.

Build

Requires monasca-common from https://opendev.org/openstack/monasca-common. Download and build following instructions in its README.rst. Then build monasca-persister by:

mvn clean package

Configuration

A sample configuration file is available in java/src/deb/etc/persister-config.yml-sample.

A second configuration file is provided in java/src/main/resources/persister-config.yml for use with the vagrant "mini-mon" development environment.

TODO

The following list is historic. Current work is tracked in Storyboard.

  • Purge metrics on shutdown
  • Add more robust offset management in Kafka. Currently, the offset is advanced as each message is read. If the Persister stops after the metric has been read and prior to it being committed to the Metrics and Alarms database, the metric will be lost.
  • Add better handling of SQL exceptions.
  • Complete health check.
  • Specify and document the names of the metrics that are available for monitoring of the Persister.
  • Document the yaml configuration parameters.