7.17.1

Prelude

Release on: 2020-02-20

• Please refer to the 7.17.1 tag on
  integrations-core
  for the list of changes on the Core Checks

Bug Fixes

• Fix a panic in the log agent when the auto-discovery reports new
  containers to monitor and the agent fails to connect to the docker
  daemon. The main setup where this happened is on ECS Fargate where
  the ECS auto-discovery is watching for new containers and the docker
  socket is not available from the datadog agent.
• Support DNS Resolution in Network Performance Monitoring when the system-probe is not
  running in a container on a non-host network.

6.17.1

6.17.1 ships the same features as 7.17.1 except for the Python versions it supports.

Please refer to the 7.17.1 changelog.

7.17.0

Prelude

Release on: 2020-02-04

• Please refer to the 7.17.0 tag on integrations-core for the list of changes on the Core Checks

Upgrade Notes

• Change agents base images to Debian bullseye

• Starting with this version, the containerized Agent never chooses
  the OS hostname as its hostname when it is running in a dedicated
  UTS namespace. This is done in order to avoid picking container IDs
  or kubernetes POD names as hostnames, since these identifiers do not
  reflect the identity of the host they run on.

  This change only affects you if your agent is currently using a
  container ID or a kubernetes POD name as hostname. The hostname of
  the agent can be checked with agent hostname. If the output stays
  stable when the container or POD of the agent is destroyed and
  recreated, you’re not impacted by this change. If the output
  changes, it means that the agent was unable to talk to EC2/GKE
  metadata server, it was unable to get the k8s node name from the
  kubelet, it was unable to get the hostname from the docker daemon
  and it is running in its dedicated UTS namespace. Under those
  conditions, you should set explicitly define the host name to be
  used by the agent in its configuration file.

New Features

• Add logic to support querying the kubelet through the APIServer to
  monitor AWS Fargate on Amazon EKS.
• Add mapping feature to dogstatsd to convert parts of
  dogstatsd/statsd metric names to tags using mapping rules in
  dogstatsd using wildcard and regex patterns.
• Resource tag collection on ECS.
• Add container_mode in journald input to set source/service as
  Docker image short name when we receive container logs

Enhancement Notes

• Add kube_node_role tag in host metadata for the node role based on
  the kubernetes.io/role label.
• Add cluster_name tag in host metadata tags. Cluster name used is
  read from config if set by user or autodiscovered from cloud
  provider or Kubernetes node label.
• The Agent check command displays the distribution metrics. The Agent
  status command displays histogram bucket samples.
• The system-probe will augment network connection information with
  DNS names gathered by inspecting local DNS traffic.
• Users can now use references to capture groups in mask sequence
  replacement_placeholder strings
• Do not apply the metric namespace configured under statsd_metric_namespace to dogstatsd
  metrics prefixed with datadog.tracer.
  Tracer metrics are published with this prefix.

Bug Fixes

• APM: The trace-agent now correctly applieslog_to_console, log_to_syslog and all other syslog
  settings.
• Make the log agent continuously retry to connect to docker rather
  than giving up when docker is not running when the agent is started.
  This is to handle the case where the agent is started while the
  docker daemon is stopped and the docker daemon is started later
  while the datadog agent is already running.
• Fixes #4650 [v7] Syntax in /readsecret.py for Py3
• Fixes an issue in Docker where mounting empty directories to disable
  docker check results in an error.
• Fixes the matching of container id in Tagger (due to runtime prefix)
  by matching on the 'id' part only
• Fix the node roles to host tags feature by handling the other
  official Kube way to setting node roles (when multiple roles are
  required)
• Properly check for "true" value of env var DD_LEADER_ELECTION
• It's possible now to reduce the risk of missing kubernetes tags on
  initial logs by configuring "logs_config.tagger_warmup_duration".
  Configuring "logs_config.tagger_warmup_duration" delays the send
  of the first logs of a container. Default value 0 seconds, the fix
  is disabled by default. Setting
  "logs_config.tagger_warmup_duration" to 5 (seconds) should be
  enough to retrieve all the tags.
• Fix eBPF code compilation errors about asm goto on Ubuntu 19.04
  (Disco)
• Fix race condition in singleton initialization
• On Windows, fixes registration of agent as event log source. Allows
  agent to correctly write to the Windows event log.
• On Windows, when upgrading, installer will fail if the user attempts
  to assign a configuration file directory or binary directory that is
  different from the original.
• Add logic to support docker restart of containers.
• Fix a Network Performance Monitoring issue where TCP connection
  direction was incorrectly classified as outgoing in containerized
  environments.
• Fixed a few edge cases that could lead to events payloads being
  rejected by Datadog's intake for being too big.

Other Notes

• Upgrade embedded dependencies: curl to 7.66.0, autoconf to
  2.69, procps to 3.3.16
• JMXFetch upgraded to
  0.34.0
• Bump embedded Python 3 to 3.7.6

6.17.0

6.17.0 ships the same features as 7.17.0 except for the Python versions it supports.

Please refer to the 7.17.0 changelog.

7.16.1

Prelude

Release on: 2020-01-06

Please refer to the 7.16.1 tag on integrations-core for the list of changes on the Core Checks

Security Issues

• UnixODBC software dependency bumped to 2.3.7 to address CVE-2018-7409.

6.16.1

Prelude

Release on: 2020-01-06

Please refer to the 7.16.1 tag on integrations-core for the list of changes on the Core Checks

Security Issues

• UnixODBC software dependency bumped to 2.3.7 to address CVE-2018-7409.

7.16.0

Prelude

Release on: 2019-12-18

This release introduces major version 7 of the Datadog Agent, which starts at v7.16.0. The only change from Agent v6 is that v7 defaults to Python 3 and only includes support for Python 3. Before upgrading to v7,
confirm that any custom checks you have are compatible with Python 3.
See this guide for more information.

Except for the supported Python versions, v7.16.0 and v6.16.0 have the same features.

Please refer to the 7.16.0 tag on integrations-core for the list of changes on the Core Checks

New Features

• Add support for SysVInit on SUSE 11.
• Add information on endpoints inside the logs-agent section of the agent status.

Enhancement Notes

• Add Python 3 linter results to status page

• Log a warning when the hostname defined in the configuration will not be used as the in-app hostname.

• Add ignore_autodiscovery_tags parameter config check.

  In some cases, a check should not receive tags coming from the autodiscovery listeners. By default ignore_autodiscovery_tags is set to false which doesn't change the behavior of the checks. The first check that will use it is kubernetes_state.

• Adds a new flare_stripped_keys config setting to clean up additional configuration information from flare.

• Adding a new config option exclude_gce_tags, to configure which metadata attribute from Google Cloud Engine to exclude from being converted into host tags.

• Extends the docker and containerd checks to include an open file descriptors metric. This metric reports the number of open file descriptors per container.

• Allow the Agent to schedule different checks from different sources on the same service.

Bug Fixes

• APM: Added a fallback into the SQL obfuscator to handle SQL engines that treat backslashes literally.
• The default list of sensitive keywords for process argument scrubbing now uses wildcards before and after.
• On Windows process agent, fix problem wherein if the agent is unable to figure out the process user name, the process info/stats were not sent at all. Now sends all relevant stats without the username
• On windows, correctly deletes python 3 precompiled files (.pyc) in the event of an installation failure and rollback
• Logs: tailed files discovered through a configuration entry with wildcard will properly have the dirname tag on all log entries.
• Fix small memory leak in datadog_agent.set_external_tags when an empty source_type dict is passed for a given hostname.
• Carry a custom patch for jaydebeapi to support latest jpype.
• Check that cluster-name provided by configuraiton file are compliant with the same rule as on GKE. Logs an error and ignore it otherwise.

Other Notes

• JMXFetch upgraded to 0.33.1
• JQuery, used in the web base agent GUI, has been upgraded to 3.4.1

6.16.0

6.16.0 ships the same features as 7.16.0 except for the Python versions it supports. Please refer to the 7.16.0 changelog.

7.15.1-beta.1

Prelude

This release is not available on Windows nor MacOS.

Release on: 2019-11-27

New Features

• Collect IP address from containers in awsvpc mode

Bug Fixes

• Reintroduce legacy checks directory to make legacy AgentCheck import path
  (from checks import AgentCheck) work again.

• Systemd integration points are re-ordered so that dbus is used in
  preference to the systemd private API at /run/systemd/private, as per
  the systemd documentation. This prevents unnecessary logging to the system
  journal when datadog-agent is run without root permissions.

6.15.1

Prelude

Release on: 2019-11-27

This release was published for Windows on 2019-12-09.

New Features

• Collect IP address from containers in awsvpc mode

Bug Fixes

• Reintroduce legacy checks directory to make legacy AgentCheck import path
  (from checks import AgentCheck) work again.

• Systemd integration points are re-ordered so that dbus is used in
  preference to the systemd private API at /run/systemd/private, as per
  the systemd documentation. This prevents unnecessary logging to the system
  journal when datadog-agent is run without root permissions.

Other Notes

• Bump embedded Python to 2.7.17.