github DataDog/datadog-agent 7.25.0


Release on: 2021-01-14

New Features

  • Add container
    auto-discovery label in AWS Fargate environment.
  • Package the gstatus command line tool binary for GlusterFS
    integration metric collection.
  • Queried domain can be tracked as part of DNS stats
  • APM: The agent is now able to skip top-level span computation in
    cases when the client has marked them by means of the
    Datadog-Client-Computed-Top-Level header.
  • APM: The maximum allowed key length for tags has been increased from
    100 to 200.
  • APM: Improve Oracle SQL obfuscation support.
  • APM: Added support for Windows pipes. To enable it, set the pipe
    path using DD_APM_WINDOWS_PIPE_NAME. For more details check PR

  • Pause containers are now detected and auto excluded based on the
    io.kubernetes container labels.
  • APM: new datadog_agent.obfuscate_sql_exec_plan
    function exposed to python checks to enable obfuscation of
    json-encoded SQL Query Execution Plans.
  • APM: new obfuscate_sql_values
    option in apm_config.obfuscation
    enabling optional obfuscation of SQL queries contained in JSON data
    collected from some APM services (ES & Mongo)

Enhancement Notes

  • Support the site option in the Windows GUI installer.
  • Adds config setting for ECS metadata endpoint client timeout
    (ecs_metadata_timeout), value in milliseconds.
  • Add loader config to allow selecting
    specific loader at runtime. This config is available at init_config and instances level.
  • Added additional container information to the status page when
    collect all container logs is enabled in agent status.
  • On Windows, it will no longer be required to supply the ddagentuser
    name on upgrade. Previously, if a non-default or domain user was
    used, the same user had to be provided on subsequent upgrades.
  • Added --flare flag to agent check to save check results to the
    agent logs directory. This enables flare to pick up check results.
  • Added new config option for JMXFetch collect_default_jvm_metrics
    that enables/disables default JVM metric collection.
  • Allow empty message for DogStatsD events (e.g. "_e{10,0}:test
  • Expires the cache key for availability of ECS metadata endpoint used
    to fetch EC2 resource tags every 5 minutes.
  • Data coming from kubernetes pods now have new kube_ownerref_kind
    and kube_ownerref_name tags for each of the pod's OwnerRef
    property, indicating its Kind and Name, respectively.
  • We improved the way Agents get the Kubernetes cluster ID from the
    Cluster Agent. It used to be that the cluster agent would create a
    configmap which had to be mounted as an env variable in the agent
    daemonset, blocking the process-agent from starting if not found.
    Now the process-agent will start, only the Kubernetes Resources
    collection will be blocked.
  • Events sent by the runtime security agent to the backend use a new
  • Scrub container args as well for orchestrator explorer.
  • Support custom autodiscovery identifiers on Kubernetes using the
    pod annotation.
  • The CPU check now collects system-wide context switches on Linux.
  • Add --table option to agent check command to output results in
    condensed tabular format instead of JSON.
  • APM: improve performance by changing the msgpack serialization
  • APM: improve the performance of the msgpack deserialization for the
    v0.5 payload format.
  • APM: improve performance of trace processing by removing some heap
  • APM: improve sublayer computation performance by reducing the number
    of heap allocations.
  • APM: improved stats computation performance by removing some string
  • APM: improved trace signature computation by avoiding heap
  • APM: improve stats computation performance.
  • Update from alpine:3.10 to alpine:3.12 the base image in Dogstatsd's

Deprecation Notes

  • APM: remove the already deprecated apm_config.extra_aggregators
    config option.

Bug Fixes

  • Fix macos dlopen failures by ensuring
    cmake preserves the required runtime search path.
  • Fix memory leak on check unscheduling, which could be noticeable for
    checks submitting large amounts of metrics/tags.
  • Exclude pause containers using the cdk/pause.* image.
  • Fixed missing some Agent environment variables in the flare
  • Fix a bug that prevented the logs Agent from discovering the correct
    init containers source and service on Kubernetes.
  • The logs agent now uses the container image name as logs source
    instead of kubernetes when a standard
    service value was defined for the container.
  • Fixes panic on concurrent map access in Kubernetes metadata tag
  • Fixed a bug that could potentially cause missing container tags for
    check metrics.
  • Fix a potential panic on ECS when the ECS API is returning empty
    docker ID
  • Fix systemd check id to handle multiple instances. The fix will make
    check id unique for each different instances.
  • Fix missing tags on pods that were not seen with a running container
  • Fix snmp listener subnet loop to use correct subnet pointer when
    creating snmpJob object.
  • Upgrade the embedded pip version to 20.3.3 to get a newer vendored
    version of urllib3.

Other Notes

  • The Agent, Logs Agent and the system-probe are now compiled with Go
  • Upgrade embedded libkrb5 Kerberos library to v1.18.3. This version
    drops support for the encryption types marked as "weak" in the docs
    of the
latest releases: 7.26.0-dbm-beta-0.5, 6.26.0, 7.26.0...
one month ago