v3.7.1

• Updating monitor collectd/haproxy to latest version

v3.7.0

• Tweaks to the bundled Varnish library to make it have the right
  dependencies. Varnish is not a first class monitor but we do bundle
  the collectd plugin for it that can be used with the collectd/custom
  monitor.
• All monitors that are wrappers for collectd Python plugins now are
  run in our own custom Python runner instead of through collectd. This
  results in a slight increase in memory (~6MB per plugin instance) since
  each plugin runs in a separate process, but they can now be restarted
  independently by the agent.

v3.6.2

• Allow MY_NODE_NAME to be referenced in config as an envvar

Breaking Changes: none

v3.6.1

• Allowing multiple monitors per port when configuring Docker container via labels
  using the docker observer.

Breaking Changes:

• None

v3.6.0

• Added the ability to configure monitors from Docker labels
• Added new whitelist to the various distributions. This is part of
  changes around host-based pricing.

Breaking Changes:

• For new installations/deployments that include the whitelist, several
  metrics that were previously sent by default will now be filtered out.
  Existing deployments that are upgrading should not be affected since
  the supplied agent config file is not overwritten on upgrades without
  manual intervention. The monitor docs now include a column for
  metrics that specify whether the metric is whitelisted.

v3.5.0

• Added environment variable remote config source
• The docker observer now supports using the hostname of the
  container in service discovery, as well as the ability to map container
  labels to dimensions on metrics that use the discovered container
  endpoint. This observer also will now discover endpoints based on
  the exposed ports of the container to better support non-standard
  network plugins to Docker.

Breaking Changes:

• The config option pollIntervalSeconds has been removed from the
  docker observer since the observer now uses the Docker API event
  stream to pick up changes instead of polling the API's container list.

v3.4.3

• Fixed a memory leak around Kubernetes pod dimension property caching.

v3.4.2

• fix issue preventing the dogstatsd listener from starting

v3.4.1

• collectd/genericjmx monitor:
  • New config options:
    • CustomDimension - takes key-value pairs that will be passed in as dimensions
    • mBeansToOmit - takes a list of MBeans to exclude
    • dimensions - takes a list of MBean properties to be passed in as dimensions
• collectd/kafka monitor:
  • Few more default metrics added
  • New required clusterName config option which will be passed in as cluster dimension
• New monitors collectd/kafka_consumer and collectd/kafka_producer for Java based Kafka consumers and producers
• Fix to RPM packaging to prevent agent.yaml config from being overwritten
• Agent will now shutdown forcibly after 10 seconds of receiving a shutdown signal

v3.4.0

• Monitor kubernetes-cluster: Fixing scaling of CPU limit/request values to always report in millicores
• Fixing RPM packaging scripts to ensure agent is enabled and restarted on upgrade on non-systemd distros
• Monitor collectd/hadoop: Upgrade collectd plugin to latest version