v3.1.1

 - Now sending K8s volume source dimensions in kubernetes-volumes monitor
 - Fixing bug where collectd/df's ignoreSelected option was always true
 - Enhancing docs for Docker observer to clarify permissions

v3.1.0

3.1.0: New Monitors and Monitor Conifg Filtering

 - New Docker container monitor to replace the old collectd one
 - New Prometheus monitor to get metrics from exporters
 - New Kubernetes Volumes monitor
 - Fixing filtering bug when reloading config in running agent
 - Fixing issue where manual JMX endpoints without name produce <nil> in template

v3.0.2

 - Adding option to use basic hostname from system instead of FQDN
   hostname

 - Filtering by k8s node when requesting pods in k8s-api observer

v3.0.1

 - Adding puppet/chef deployment modules

 - Fixing k8s secret annotation bug
 - Adding umount script for k8s to fix pod stuck issue
 - Making collectd custom monitors run in a separate collectd instance so
   that we can treat them exactly the same as other monitors otherwise
 - Making config watch configurable in config file itself instead of via
   CLI args

v3.0.0

This release introduces a completely new Docker scratch image base that is much smaller than before (down to 150MB from around 270MB before). It is intended to almost exactly mirror the agent as it will run outside of Docker so that we have parity between the two.

Also included is Kubernetes annotation config support (see https://github.com/signalfx/signalfx-agent/tree/master/docs/kubernetes-setup.md).

If you are manually configuring a monitor, the old serviceEndpoints config option has been merged with the top level monitor config (e.g. just put host and port directly on the monitor config object. You can only specify one custom endpoint per monitor config instance, but you can use YAML anchors/references to reduce duplication if you have a lot of manually specified hosts.

All of the collectd plugins that were previously configured implicitly are now explicitly configured in the monitor config. Be sure to add them or else you will lose several host-level metrics.

The procFSPath config option has been removed in favor of specifying it in the monitors that need it. This further decouples the monitors from the agent core for easier maintainability.

Discovery rule variables have changed from camelCased to snake_case so that they are consistent with our dimension names. Any dimensions that are added to a discovered endpoint can now be used in discovery rules. For example, you can now match on kubernetes_pod_name. The old camelCased variables are retained for backwards compatibility, but it is better to convert now.

v2.1.0: New Monitors

Added monitors for HAProxy, Consul, Etcd, and Spark, as well as a new monitor that will fetch cAdvisor metrics through the kubelet API instead of directly from cAdvisor.

Also simplified manually defined service endpoints such that they cannot include config that will override monitor config anymore. That was too complicated and not worth it.

v2.0.0: Config Overhaul

A total rewrite of the agent to use a single config file and to support manually defined service endpoints.

v1.0.0

• Initial release of the Signal Fx Agent

v1.0.0-beta-8

• Fixed issue preventing the agent from starting in some environments

v1.0.0-beta-6

Merge pull request #136 from signalfx/rabbitmq

Rabbitmq