v3.9.3

• Allow monitor kubernetes-cluster to be restricted to a specific namespace
• Allow observer k8s-api to be restricted to a specific namespace
• Add Kubernetes leader agent information to the status output as well as a new internal metric

Breaking Changes: None

v3.9.2

• New metricsToInclude config option to make allowing through custom metrics from standard monitors easier (see https://github.com/signalfx/signalfx-agent/blob/master/docs/filtering.md#inclusion-filtering)
• Updating the underlying collectd plugin for monitor collectd/jenkins to 1.0.6 for better error handling

Breaking Changes: none

v3.9.1

• Updating filtering to automatically combine filters that share the same monitorType
• Enhancing monitor kubernetes-cluster to sync properties related to Deployments to the kubernetes_pod_uid dimension

Breaking Changes: None

v3.9.0

• Monitor docker-container-stats: Add option to map envvars on container to dimensions
• Monitor docker-container-stats: Add docker_hostname dimension to all datapoints that is the container's hostname
• Monitor collectd/elasticsearch: Fix bug with conversion of list config options

Breaking Changes:

• docker-container-stats MTSs will be recreated due to the addition of the new docker_hostname dimension

v3.8.4

• Some bugfixes for monitor conviva that will reduce DPM significantly
• The number of trace spans in flight is now capped at a configurable upper limit to prevent memory exhaustion

Breaking Changes: None

v3.8.3

• Adding monitor conviva to pull metrics from the Conviva streaming platform
• Fixing a bug in the trace writer to handle errors without locking up
• Adding metric for dropped trace spans

Breaking Changes: None

v3.8.2

• Monitor collectd/docker: Update underlying plugin to version 3.0.0

Breaking Changes: None

v3.8.1

• Added dimension property filtering

Breaking Changes: None

v3.8.0

• Added cluster option to collectd/elasticsearch monitor
• docker-container-stats monitor's memory.percent metric now excludes page cache usage
• The host observer's poll interval is now configurable
• Added support for trace span forwarding via the trace-forwarder monitor
• IPv6 is temporarily disabled in the host observer until we figure out how to best handle that
• The docker observer now has an option to use host-bound ports instead of the docker network host/port

Breaking Changes:

• The memory.percent metric from docker-container-stats will probably be consistently lower than before, but MTSs should remain consistent
• IPv6 services will no longer be discovered by the host observer

v3.7.2

• Fixed regex pattern for disks exclusion pattern in collectd/disk monitor to conform to POSIX extended RE
• Fixed race condition in the Python runner where metrics could be emitted before configuration was completed

Breaking Changes: None