Kubernetes Controller Manager

Supported OS Windows Mac OS

통합 버전7.1.0

Kube Controller Manager 대시보드

개요

이 점검은 Kubernetes 컨트롤 플레인의 일부인 Kubernetes Controller Manager를 모니터링합니다.

참고: 이 점검은 해당 서비스가 노출되지 않기 때문에 Amazon EKS 클러스터에 대한 데이터를 수집하지 않습니다.

설정

설치

Kubernetes Controller Manager 점검은 Datadog Agent 패키지에 포함되어 있으므로 서버에 추가 설치할 필요가 없습니다.

구성

  1. Agent의 설정 디렉터리 루트에서 conf.d/ 폴더에 있는 kube_controller_manager.d/conf.yaml 파일을 편집하여 kube_controller_manager 성능 데이터 수집을 시작합니다. 사용 가능한 모든 설정 옵션은 샘플 kube_controller_manager.d/conf.yaml을 참조하세요.

  2. Agent를 재시작합니다

이 통합을 위해서는 컨트롤러 관리자의 메트릭 엔드포인트에 대한 액세스가 필요합니다. 메트릭 엔드포인트에 액세스하려면 다음을 수행해야 합니다.

  • controller-manager 프로세스의 IP/포트에 대한 액세스가 있어야 합니다.
  • /metrics 엔드포인트에 대한 get RBAC 권한이 있어야 합니다. (기본 Datadog Helm 차트에는 이미 이에 대한 올바른 RBAC 역할과 바인딩이 추가되어 있습니다.)

검증

Agent의 status 하위 명령을 실행하고 Checks 섹션에서 kube_controller_manager를 찾습니다.

수집한 데이터

메트릭

kube_controller_manager.goroutines
(gauge)
Number of goroutines that currently exist
kube_controller_manager.job_controller.terminated_pods_tracking_finalizer
(count)
Used to monitor whether the job controller is removing Pod finalizers from terminated Pods after accounting them in Job status
kube_controller_manager.leader_election.lease_duration
(gauge)
Duration of the leadership lease
kube_controller_manager.leader_election.transitions
(count)
Number of leadership transitions observed
kube_controller_manager.max_fds
(gauge)
Maximum allowed open file descriptors
kube_controller_manager.nodes.count
(gauge)
Number of registered nodes, per zone
kube_controller_manager.nodes.evictions
(count)
Count of node eviction events, per zone
kube_controller_manager.nodes.unhealthy
(gauge)
Number of unhealthy nodes, per zone
kube_controller_manager.open_fds
(gauge)
Number of open file descriptors
kube_controller_manager.queue.adds
(count)
Elements added, by queue
kube_controller_manager.queue.depth
(gauge)
Current depth, by queue
kube_controller_manager.queue.latency.count
(gauge)
Processing latency count, by queue (deprecated in kubernetes v1.14)
kube_controller_manager.queue.latency.quantile
(gauge)
Processing latency quantiles, by queue (deprecated in kubernetes v1.14)
Shown as microsecond
kube_controller_manager.queue.latency.sum
(gauge)
Processing latency sum, by queue (deprecated in kubernetes v1.14)
Shown as microsecond
kube_controller_manager.queue.process_duration.count
(gauge)
How long processing an item from workqueue takes, by queue
kube_controller_manager.queue.process_duration.sum
(gauge)
Total workqueue processing time, by queue
Shown as second
kube_controller_manager.queue.queue_duration.count
(gauge)
How long item stays in a queue before being requested, by queue
kube_controller_manager.queue.queue_duration.sum
(gauge)
Total time of items stays in a queue before being requested, by queue
Shown as second
kube_controller_manager.queue.retries
(count)
Retries handled, by queue
kube_controller_manager.queue.work_duration.count
(gauge)
Work duration, by queue (deprecated in kubernetes v1.14)
kube_controller_manager.queue.work_duration.quantile
(gauge)
Work duration quantiles, by queue (deprecated in kubernetes v1.14)
Shown as microsecond
kube_controller_manager.queue.work_duration.sum
(gauge)
Work duration sum, by queue (deprecated in kubernetes v1.14)
Shown as microsecond
kube_controller_manager.queue.work_longest_duration
(gauge)
How many seconds has the longest running processor been running, by queue
Shown as second
kube_controller_manager.queue.work_unfinished_duration
(gauge)
How many seconds of work has done that is in progress and hasn't been observed by process_duration, by queue
Shown as second
kube_controller_manager.rate_limiter.use
(gauge)
Usage of the rate limiter, by limiter
kube_controller_manager.slis.kubernetes_healthcheck
(gauge)
Result of a single controller manager healthcheck (alpha; requires k8s v1.26+)
kube_controller_manager.slis.kubernetes_healthcheck_total
(count)
Cumulative results of all controller manager healthchecks (alpha; requires k8s v1.26+)
kube_controller_manager.threads
(gauge)
Number of OS threads created

이벤트

Kubernetes Controller Manager 점검은 이벤트를 포함하지 않습니다.

서비스 점검

kube_controller_manager.prometheus.health
Returns CRITICAL if the check cannot access the metrics endpoint.
Statuses: ok, critical

kube_controller_manager.leader_election.status
Returns CRITICAL if no replica is currently set as leader.
Statuses: ok, critical

kube_controller_manager.up
Returns CRITICAL if Kube Controller Manager is not healthy.
Statuses: ok, critical

트러블슈팅

도움이 필요하신가요? Datadog 지원팀에 문의하세요.

PREVIEWING: guacbot/translation-pipeline