v3.3.2

• upgrade to latest spark plugin
• add collectd/hadoop monitor
• add collectd/hadoopjmx monitor

v3.3.1

• Cap the size of dimension property deduplication history set to prevent unbounded memory usage with large number of dimension values
• Make dimension property update requests in parallel, up to a configurable max limit

Breaking Changes: none

v3.3.0

• Monitor kubelet-stats: Ensures that a datapoint is sent every interval, even if kubelet doesn't have a new datapoint available. It will always send the last datapoint gotten from kubelet, even if it was the same as before. This should prevent the need for using extrapolation in charts as much.

• New monitor host-metadata: this replaces functionality in the collectd/signalfx-metadata monitor that adds properties about the kernel and machine to the 'host' dimension. It is included in the config for new installs, but you must add this monitor to your agent config if you want host metadata.

• The Get function in discovery rules now returns nil if the key does not
  exist in the map. Previously it would raise an error that would cause the
  rule to always fail.

• The way datapoints are sent has been reworked so that datapoints are sent with less delay and potentially in parallel. This should help reduce lag over high-latency connections. As a consequence, the config options writer.datapointBufferCapacity, writer.datapointBufferHardMax, and writer.datapointSendIntervalSeconds have been removed and the options writer.datapointMaxBatchSize and writer.datapointMaxRequests have been added. See the config schema doc for more details.

• Monitor kubernetes-cluster: New resource quota metrics to show namespace resource quotas. Note that you have to add that resource to the ClusterRole for the agent to give it permission to fetch those from the API Server.

• Monitor kubernetes-cluster: No longer send the host dimension for kubernetes node and pod phase metrics. This will break some MTSs but will prevent confusion when that host name doesn't match the same (potentially fully-qualified) hostname from other metrics.

• Monitor collectd/genericjmx: Allow passing through the InstanceFrom option to collectd.

• Optimizations to the JSON decoding of the collectd metric receiver in the agent may reduce CPU usage of the agent when collectd is sending a lot of metrics.

Breaking Changes:

• Certain metrics from the collectd/activemq monitor that already had instanceFrom specified on their default mbean definitions will now have a different plugin_instance dimension, which will break some MTSs.

• The kubernetes.node_ready metric is losing the host dimension and will cause new MTSs to be created.

• The kubernetes.pod_phase, kubernetes.container_restart_count, and kubernetes.container_ready metrics have the host dimension renamed to kubernetes_node, which is what it should have been all along. This will cause new MTSs to be formed for those metrics.

v3.2.4

 - Monitor collectd/kong: added
 - Monitor collectd/chrony: added
 - Fixed bug where raw remote values with wildcards were not converted
   to string correctly

v3.2.3

 - Monitor collectd/health-checker: Update health_checker plugin to
   latest version and have better config

v3.2.2

 - Fixed bug related to Get() function in discovery rules
 - Added examples for Get() and Contains() to auto-discovery documentation

v3.2.1

 - improve error handling in signalfx metadata plugin
 - upgrade couchbase plugin to latest version

v3.2.0

 - collectd/mysql monitor: Add support for MySQL 8.x and drop support for 4.x.

 - Added support for using node name to unique identify Kubernetes
   nodes instead of machine id.  Machine id proved too unreliable across
   various K8s deployments, including EKS and PKS.  It is still the
   default, but an error message will be shown if it is duplicated or
   missing.  The use of node name is documented in the K8s setup guide
   under the section for EKS.  Existing configurations will behave the
   same as before

Breaking Changes:

 - MySQL 4.x is no longer supported

v3.1.13

- add collectd/jenkins monitor
- make skipVerify kubelet client option default to true
- decrease default agent interval to 10 seconds from 15 seconds
- ensure all existing monitors are at their latest code release

v3.1.12

- Update metadata plugin configuration options to better match standalone collectd
- Add k8s integration tests
- Wait for services to launch before executing integration tests