6.6.0
Download links
Changes
Prelude
Release on: 2018-10-25
- Please refer to the 6.6.0 tag on integrations-core for the list of changes on the Core Checks.
- Please refer to the 6.6.0 tag on trace-agent for the list of changes on the Trace Agent.
- Please refer to the 6.6.0 tag on process-agent for the list of changes on the Process Agent.
New Features
- Disk check support for the puppy agent on unix-like systems
- Support for the upcoming cluster-agent cluster-level checks feature,
via theclusterchecks
config provider - Add a new CRI core check that will send metrics about resource usage
of your containers via the Container Runtime Interface. - Support SysVinit on Debian note: some warnings can appear if you
enable/disable the agent manually on a systemd system. They can be
safely ignored - The
datadog-agent integration install
command will now check for
compatibility withdatadog-checks-base
shipped with the agent. In
case of mismatch, it will try to rollback to the previously
installed integration version and exit with a failure. - Add
--in-toto
flag todatadog-agent integration
command to
enable in-toto - Add
--verbose
flag todatadog-agent integration
command to
enable verbose logging on pip and TUF - Docker image: running with a read-only root filesystem is now
supported
Enhancement Notes
- Add a setting to configure the interval at which configs should be
polled for autodiscovery. - Support a new config option,
site
, that allows setting the Datadog
site to which the Agent should send data.dd_url
is still
supported and, when set, overridessite
. - Display a warning in the agent status when too many logs are being
tailed and the agent is not tailing them all. This happens with
wildcards in path of the tailed files - Dogstatsd supports removing the hostname on events and services
checks as it did with metrics, by adding an emptyhost:
tag - Added new dogstatsd_tags variable which can be used to specify
additional tags to append to all metrics received by dogstatsd. - dogstatsd cleans up stale UNIX socket on startup.
- The ecs-agent's docker container name can now be set via the
ecs_agent_container_name
option or the
DD_ECS_AGENT_CONTAINER_NAME
envvar for autodetection. - EKS pause containers are ignored by default
- All python and go checks support the new
empty_default_hostname
option to send metrics with no hostname. This is used for
cluster-level checks - All go checks now support the
min_collection_interval
option, as
python check already do - Added a
kubelet_wait_on_missing_container
option to handle hosts
where the kubelet's podlist is slow to update, leading to missing
tags or failing Autodiscovery. Set it to 1 for a 1 second maximum
wait - Add an option to enable protobuf communication with the Kubernetes
apiserver datadog-agent integration
command will not pull any of the
integration's dependencies- More accurate tag extraction logic for Docker Swarm
- Added new command line properties to the Windows installer which
allow for setting site specific configuration.
Bug Fixes
- Fix an issue preventing the exit logs of the agent from displaying
the correct filename. - Fix bug that occurs when checks labels/annotation are misconfigured
and would prevent the logs of the container to be tailed - Fix an issue causing the agent to stop when systemd-journald service
is stopped or fails - Fix deadlock when an config item under
logs
is invalid - Fix system.mem.pct_usable implementation on Linux 3.14+ to match
Datadog Agent 5 - Fix a potential race in the autodiscovery where a service would be
removed before its config could be resolved (causing the agent to
crash) - Fixes crash on Windows when the agent encounters a malformed
performance counter database - Fixes config.Digest that was not stable depending on the oder of
tags in the instance. It also did not take into account LogsConfig,
this is fixed as well. - Fix an issue where the log agent would prevent files from being log
rotated on Windows - Correctly pass the agent's proxy settings to pip when using the
datadog-agent integration
command with TUF enabled. - Recover from errors when connection to the docker socket is lost to
continue tailing containers. - When installing / updating wheels using the
datadog-agent integration
command, we replace the PyPI index with
our own by default, in order to prevent accidental installation of
Datadog or even third-party packages from PyPI. - Remove some undocumented power user options to the
datadog-agent integration
command to prevent accidental
misconfiguration that may reduce security guarantees.
Other Notes
- JMXFetch upgraded to 0.21.0; Adds support for rmi registry
connection over SSL and client authentication. - Use autodiscovery in log-agent kubernetes integration