misterli's Blog.

prometheus使用missing-container-metrics监控pod oomkill

prometheus oomkill
字数统计: 1.7k阅读时长: 8 min
2021/07/06

简介

Kubernetes 默认情况下使用 cAdvisor 来收集容器的各项指标，足以满足大多数人的需求，但还是有所欠缺，比如缺少对以下几个指标的收集：

  • OOM kill
  • 容器重启的次数
  • 容器的退出码

missing-container-metrics 这个项目弥补了 cAdvisor 的缺陷，新增了以上几个指标，集群管理员可以利用这些指标迅速定位某些故障。例如，假设某个容器有多个子进程，其中某个子进程被 OOM kill，但容器还在运行，如果不对 OOM kill 进行监控，管理员很难对故障进行定位。

安装

官方提供了helm chart方式来进行安装，我们先添加helm仓库

1
helm repo add missing-container-metrics https://draganm.github.io/missing-container-metrics

把这个chart下载到本地，我们需要修改value.yaml文件

1
2
3
4
5
[root@master-01 addons]# helm pull missing-container-metrics/missing-container-metrics
[root@master-01 addons]# ls
blackbox        dingtalk  harbor_exporter                      mysql-exporter    prometheusalert  rules                          servicemonitor  victoriametrics
blackbox-probe  etcd      missing-container-metrics-0.1.1.tgz  process-exporter  redis-exporter   scheduler-controller-svc.yaml  ssl-exporter
[root@master-01 addons]# tar xf  missing-container-metrics-0.1.1.tgz

可配置项

Parameter Description Default
image.repository 镜像名称 dmilhdef/missing-container-metrics
image.pullPolicy 镜像拉取策略 IfNotPresent
image.tag 镜像tag v0.21.0
imagePullSecrets 拉取镜像的secret []
nameOverride 覆盖生成的图表名称。默认为 .Chart.Name。
fullnameOverride 覆盖生成的版本名称。默认为 .Release.Name。
podAnnotations Pod 的Annotations {"prometheus.io/scrape": "true", "prometheus.io/port": "3001"}
podSecurityContext 为 pod 设置安全上下文
securityContext 为 pod 中的容器设置安全上下文
resources PU/内存资源请求/限制 {}
useDocker 从 Docker 获取容器信息,如果容器运行时为docker ,设置为true false
useContainerd 从 Containerd 获取容器信息,如果容器运行时为containers ,设置为true true

我们这里修改missing-container-metrics/values.yaml中``useDocker为 true`，然后安装

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
[root@master-01 addons]# kubectl create namespace missing-container-metrics
namespace/missing-container-metrics created
[root@master-01 addons]# helm install missing-container-metrics  missing-container-metrics -n missing-container-metrics
NAME: missing-container-metrics
LAST DEPLOYED: Tue Jul  6 10:47:35 2021
NAMESPACE: missing-container-metrics
STATUS: deployed
REVISION: 1
TEST SUITE: None
[root@master-01 addons]# helm -n missing-container-metrics list
NAME                            NAMESPACE                       REVISION        UPDATED                                 STATUS          CHART                           APP VERSION
missing-container-metrics       missing-container-metrics       1               2021-07-06 10:47:35.261058822 +0800 CST deployed        missing-container-metrics-0.1.1 0.21.0     
##因为我只有一个节点，所以这里daemonset 就只有一个pod
[root@master-01 addons]# kubectl get pod -n missing-container-metrics 
NAME                              READY   STATUS    RESTARTS   AGE
missing-container-metrics-s9cgk   1/1     Running   0          115s

我们可以访问服务的3001端口查看metrics,例如

1
2
3
4
5
6
7
8
[root@master-01 addons]# curl 100.67.79.150:3001/metrics
# HELP container_last_exit_code Last exit code of the container
# TYPE container_last_exit_code gauge
container_last_exit_code{container_id="docker://0133fb5d739ba98b3985bdc7766fa200334bbbf29de9a61f98a463ec00de53de",container_short_id="0133fb5d739b",docker_container_id="0133fb5d739ba98b3985bdc7766fa200334bbbf29de9a61f98a463ec00de53de",image_id="docker-pullable://k8s.gcr.io/pause:3.2",name="k8s_POD_dns-autoscaler-565bf94d6c-dc6v4_kube-system_96437fe8-200c-4845-a7cc-a27790c6c5a7_0",namespace="kube-system",pod="dns-autoscaler-565bf94d6c-dc6v4"} 0
container_last_exit_code{container_id="docker://0388ba15b0181fead17cfc3606a57aeef0a9b8b73cf3f97eb901565c8aa1702c",container_short_id="0388ba15b018",docker_container_id="0388ba15b0181fead17cfc3606a57aeef0a9b8b73cf3f97eb901565c8aa1702c",image_id="docker-pullable://sha256:e20d2ec0d0ed8ffd693b435af9f2943095a608440e3b845331d6d00344025455",name="k8s_victoriametrics_victoriametrics-0_kube-system_7b381d2c-791b-4e38-8cbb-43485afcb285_0",namespace="kube-system",pod="victoriametrics-0"} 0
container_last_exit_code{container_id="docker://0400f7e29dab47304f97669cb52b5c7c9310fbb5c156c07d0dc9bfca6b8ee14d",container_short_id="0400f7e29dab",docker_container_id="0400f7e29dab47304f97669cb52b5c7c9310fbb5c156c07d0dc9bfca6b8ee14d",image_id="docker-pullable://k8s.gcr.io/pause:3.2",name="k8s_POD_csi-resizer-f6d66495f-s4vkv_longhorn-system_282278da-2638-4e26-8411-802bf57c1ed8_0",namespace="longhorn-system",pod="csi-resizer-f6d66495f-s4vkv"} 0
container_last_exit_code{container_id="docker://04e2c60777ce277c62c7137f1d7b40d9c1523bb3edf9127efd357590f39ba79c",container_short_id="04e2c60777ce",docker_container_id="04e2c60777ce277c62c7137f1d7b40d9c1523bb3edf9127efd357590f39ba79c",image_id="docker-pullable://k8s.gcr.io/pause:3.2",name="k8s_POD_kube-state-metrics-859b6bf99-q8tdf_monitoring_529aa188-f7a0-4b5c-9608-cd8fc473ac8c_2",namespace="monitoring",pod="kube-state-metrics-859b6bf99-q8tdf"} 0

服务公开了如下的指标：

  • container_restarts ：容器的重启次数。
  • container_ooms ：容器的 OOM 杀死数。这涵盖了容器 cgroup 中任何进程的 OOM 终止。
  • container_last_exit_code ：容器的最后退出代码。

每一个指标包含如下标签：

  • docker_container_id：容器的完整 ID。
  • container_short_id：Docker 容器 ID 的前 6 个字节。
  • container_id：容器 id 以与 kubernetes pod 指标相同的格式表示 - 以容器运行时为前缀docker://并containerd://取决于容器运行时。这使得 Prometheus 中的kube_pod_container_info指标可以轻松连接。
  • name：容器的名称。
  • image_id：图像 id 以与 k8s pod 的指标相同的格式表示。这使得 Prometheus 中的kube_pod_container_info指标可以轻松连接。
  • pod：如果io.kubernetes.pod.name在容器上设置了pod标签，则其值将设置为指标中的标签
  • namespace：如果io.kubernetes.pod.namespace容器上设置了namespace标签，则其值将设置为指标的标签。

添加PodMonitor 和 PrometheusRule（针对Prometheus Operator）

在template目录下创建文件podmonitor.yaml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
{{ if .Values.prometheusOperator.podMonitor.enabled }}
apiVersion: monitoring.coreos.com/v1
kind: PodMonitor
metadata:
  name: {{ include "missing-container-metrics.fullname" . }}
  {{- with .Values.prometheusOperator.podMonitor.namespace }}
  namespace: {{ . }}
  {{- end }}
  labels:
    {{- include "missing-container-metrics.labels" . | nindent 4 }}
    {{- with .Values.prometheusOperator.podMonitor.selector }}
    {{- toYaml . | nindent 4 }}
    {{- end }}
spec:
  selector:
    matchLabels:
      {{- include "missing-container-metrics.selectorLabels" . | nindent 6 }}
  podMetricsEndpoints:
  - port: http
  namespaceSelector:
    matchNames:
      - {{ .Release.Namespace }}
{{ end }}

在template目录下创建文件prometheusrule.yaml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
{{ if .Values.prometheusOperator.prometheusRule.enabled }}
apiVersion: monitoring.coreos.com/v1
kind: PrometheusRule
metadata:
  name: {{ include "missing-container-metrics.fullname" . }}
  {{- with .Values.prometheusOperator.prometheusRule.namespace }}
  namespace: {{ . }}
  {{- end }}
  labels:
    {{- include "missing-container-metrics.labels" . | nindent 4 }}
    {{- with .Values.prometheusOperator.prometheusRule.selector }}
    {{- toYaml . | nindent 4 }}
    {{- end }}
spec:
  groups:
  - name: {{ include "missing-container-metrics.fullname" . }}
    rules:
      {{- toYaml .Values.prometheusOperator.prometheusRule.rules | nindent 6 }}
{{ end }}

修改value.yaml,添加如下

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
useDocker: true
useContainerd: false
###添加
prometheusOperator:
  podMonitor:
    # Create a Prometheus Operator PodMonitor resource
    enabled: true
    # Namespace defaults to the Release namespace but can be overridden
    namespace: ""
    # Additional labels to add to the PodMonitor so it matches the Operator's podMonitorSelector
    selector:
      app.kubernetes.io/name: missing-container-metrics

  prometheusRule:
    # Create a Prometheus Operator PrometheusRule resource
    enabled: true
    # Namespace defaults to the Release namespace but can be overridden
    namespace: ""
    # Additional labels to add to the PrometheusRule so it matches the Operator's ruleSelector
    selector:
      prometheus: k8s
      role: alert-rules
    # The rules can be set here. An example is defined here but can be overridden.
    rules:
    - alert: ContainerOOMObserved
      annotations:
        message: A process in this Pod has been OOMKilled due to exceeding the Kubernetes memory limit at least twice in the last 15 minutes. Look at the metrics to determine if a memory limit increase is required.
      expr: sum(increase(container_ooms[15m])) by (exported_namespace, exported_pod) > 2
      labels:
        severity: warning
    - alert: ContainerOOMObserved
      annotations:
        message: A process in this Pod has been OOMKilled due to exceeding the Kubernetes memory limit at least ten times in the last 15 minutes. Look at the metrics to determine if a memory limit increase is required.
      expr: sum(increase(container_ooms[15m])) by (exported_namespace, exported_pod) > 10
      labels:
        severity: critical

使用下面命令更新

1
2
3
4
5
6
7
8
[root@master-01 addons]# helm upgrade missing-container-metrics -n missing-container-metrics missing-container-metrics/
Release "missing-container-metrics" has been upgraded. Happy Helming!
NAME: missing-container-metrics
LAST DEPLOYED: Tue Jul  6 11:36:02 2021
NAMESPACE: missing-container-metrics
STATUS: deployed
REVISION: 2
TEST SUITE: None

更新后会创建podmonitor和prometeusrules

1
2
3
4
5
6
[root@master-01 addons]# kubectl get prometheusrules.monitoring.coreos.com  -n monitoring
NAME                        AGE
missing-container-metrics   15s
[root@master-01 addons]# kubectl get podmonitors.monitoring.coreos.com  -n missing-container-metrics 
NAME                        AGE
missing-container-metrics   35s

我们可以在prometheus ui 上看到相关target和rules

image-20210706113837427

image-20210706114210120

原文作者：Mister Li

原文链接：https://www.lishuai.fun/2021/07/06/prometheus-create-missing-containerer-mitrics/

发表日期：July 6th 2021, 11:45:37 am

更新日期：May 3rd 2022, 11:25:55 pm

版权声明：本文采用知识共享署名-非商业性使用 4.0 国际许可协议进行许可

CATALOG
  1. 1. 简介
  2. 2. 安装
  3. 3. 添加PodMonitor 和 PrometheusRule（针对Prometheus Operator）
Total : 48
2024
2023
2022
2021
2020
kubernetes rbac k8s vulnerability ingress-nginx RBAC argocd devops gitlab harbor slb hugo pv csi s3 kubesphere prometheus sso kubevela k8s准入控制 kyverno longhorn oomkill event configmap secret sync ingress cert-manager chaos-mesh minio vpa 动态扩缩容 k8s kubevela oamm 报警 ssl-exporter traefik 混沌工程 rke k8s pvc nfs prometheus redis umami etcd kuboard docker ci buildkit redis dockerhub proxy Cloudflare