k8s dashboard
的界面中仅显示了pod
一些配置信息以及node层面的监控,无法图形化展现集群度量指标信息。
下图为dashboard界面:
而如果要展示图形化的集群度量指标信息,就需要安装一个dashboard
插件:heapster
。
Heapster
原生支持K8s(v1.0.6及以后版本)
和CoreOS
,并且支持多种存储后端,比如:InfluxDB
、ElasticSearch
、Kafka
等,这个风格和k8s
的确很像:功能先不管完善与否,先让自己在各个平台能用起来再说。这里我们使用的数据存储后端是InfluxDB
。
参考文档:https://blog.csdn.net/u010955999/article/details/79425592
我们的Heapster也是要放在pod里运行的。当前,Heapster的最新stable版本是v1.2.0,我们可以下载其源码包到K8s cluster上的某个Node上。解压后,我们得到一个名为”heapster-1.2.0″的目录,进入该目录,我们可以看到如下内容:
root@node1:~/k8stest/dashboardinstall/heapster-1.2.0# ls
code-of-conduct.md CONTRIBUTING.md docs Godeps hooks integration LICENSE metrics riemann version
common deploy events grafana influxdb kafka Makefile README.md vendor
以InfluxDB为存储后端的Heapster部署yaml在deploy/kube-config/influxdb下面:
root@node1:~/k8stest/dashboardinstall/heapster-1.2.0# ls -l deploy/kube-config/influxdb/
total 28
-rw-r--r-- 1 root root 414 Sep 14 12:47 grafana-service.yaml
-rw-r--r-- 1 root root 942 Jan 20 15:15 heapster-controller.yaml
-rw-r--r-- 1 root root 249 Sep 14 12:47 heapster-service.yaml
-rw-r--r-- 1 root root 1465 Jan 19 21:39 influxdb-grafana-controller.yaml
-rw-r--r-- 1 root root 259 Sep 14 12:47 influxdb-service.yaml
这里有五个yaml(注意:与heapster源码库中最新的代码已经有所不同,最新代码将influxdb和grafana从influxdb-grafana-controller.yaml拆分开了)。其中的一些docker image在墙外,如果你有加速器,那么你可以直接执行create命令;否则最好找到一些替代品: 比如:用signalive/heapster_grafana:2.6.0-2替换gcr.io/google_containers/heapster_grafana:v2.6.0-2。
创建pod的操作很简单:
~/k8stest/dashboardinstall/heapster-1.2.0# kubectl create -f deploy/kube-config/influxdb/
service "monitoring-grafana" created
replicationcontroller "heapster" created
service "heapster" created
replicationcontroller "influxdb-grafana" created
service "monitoring-influxdb" created
如果image pull顺利的话,那么这些pod和service的启动是会很正常的。
//kube get pods -n kube-system
... ...
kube-system heapster-b1dwa 1/1 Running 0 1h 172.16.57.9 10.46.181.146 k8s-app=heapster,version=v6
kube-system influxdb-grafana-8c0e0 2/2 Running 0 1h 172.16.57.10 10.46.181.146 name=influxGrafana
... ...
我们用浏览器打开kubernetes的Dashboard,期待中的图形化和集群度量指标信息到哪里去了呢?Dashboard还是一如既往的如上面图示中那样“简朴”,显然我们遇到问题了!
问题在哪?我们需要逐个检视相关Pod的日志:
# kubectl logs -f pods/influxdb-grafana-xxxxxx influxdb -n kube-system
# kubectl logs -f pods/influxdb-grafana-xxxxxx grafana -n kube-system
# kubectl logs -f pods/heapster-xxxxx -n kube-system
在heapster-xxxxx这个pod中,我们发现了大量失败日志:
E0119 13:14:37.838900 1 reflector.go:203] k8s.io/heapster/metrics/heapster.go:319: Failed to list *api.Pod: the server has asked for the client to provide credentials (get pods)
E0119 13:14:37.838974 1 reflector.go:203] k8s.io/heapster/metrics/processors/node_autoscaling_enricher.go:100: Failed to list *api.Node: the server has asked for the client to provide credentials (get nodes)
E0119 13:14:37.839516 1 reflector.go:203] k8s.io/heapster/metrics/processors/namespace_based_enricher.go:84: Failed to list *api.Namespace: the server has asked for the client to provide credentials (get namespaces)
heapster无法连接apiserver,获取不要想要的信息。从kube-apiserver的日志(/var/log/upstart/kube-apiserver.log)也印证了这一点:
E0120 09:15:30.833928 12902 handlers.go:54] Unable to authenticate the request due to an error: crypto/rsa: verification error
E0120 09:15:30.834032 12902 handlers.go:54] Unable to authenticate the request due to an error: crypto/rsa: verification error
E0120 09:15:30.835324 12902 handlers.go:54] Unable to authenticate the request due to an error: crypto/rsa: verification error
从apiserver的日志来看,heapster是通过apiserver的secure port连接的,由于我们的API server设置有https client端证书校验机制,因此两者连接失败。
现在我们就来解决上述问题。
首先,我们会想到:能否让heapster通过kube APIServer的insecure-port连接呢?在《Kubernetes集群的安全配置》一文中我们提到过,kube-apiserver针对insecure-port接入的请求没有任何限制机制,这样heapster就可以获取到它所想获取到的所有有用信息。
在heapster doc中的“Configuring Source”中,我们找到了连接kube-apiserver insecure-port的方法。不过在修改yaml之前,我们还是要先来看看当前heapster的一些启动配置的含义:
//deploy/kube-config/influxdb/heapster-controller.yaml
command:
- /heapster
- --source=kubernetes:https://kubernetes.default
- --sink=influxdb:http://monitoring-influxdb:8086
我们看到heapster启动时有两个启动参数:
–source指示数据源,heapster是支持多种数据源的,这里用的是“kubernetes”类型的数据源,地址是:kubernetes.default。这个域名的全名是:kubernetes.default.svc.cluster.local,就是service “kubernetes”在cluster中的域名,而”kubernetes”服务就是kube-apiserver,它的信息如下:
# kubectl get services
NAME CLUSTER-IP EXTERNAL-IP PORT(S) AGE
kubernetes 10.254.0.1 443/TCP 99d
... ...
# kubectl describe svc/kubernetes
Name: kubernetes
Namespace: default
Labels: component=apiserver
provider=kubernetes
Selector:
Type: ClusterIP
IP: 10.254.0.1
Port: https 443/TCP
Endpoints: xxx.xxx.xxx.xxx:6443
Session Affinity: ClientIP
No events.
因此,该域名在k8s DNS中会被resolve为clusterip:10.254.0.1。外加https的默认端口是443,因此实际上heapster试图访问的apiserver地址是:https://192.168.3.1:443。
heapster启动的另外一个参数是–sink,这个传入的就是存储后端,我们使用了InfluxDB,这里传入的就是上面创建的InfluxDB service的域名和端口号,我们在cluster中也能查找到该Service的信息:
# kubectl get services -n kube-system
NAME CLUSTER-IP EXTERNAL-IP PORT(S) AGE
monitoring-influxdb 10.254.0.1 8083/TCP,8086/TCP 1h
... ...
前面提到过,我们的APIServer在secure port上是有client端证书校验的,那么以这样的启动参数启动的heapster连接不上kube-apiserver就“合情合理”了。
接下来,我们按照”Configuring Source”中的方法,将heapster与kube-apiserver之间的连接方式改为通过insecure port进行:
// kube-config/influxdb/heapster-controller.yaml
... ...
command:
- /heapster
- --source=kubernetes:http://192.168.201.128:8080?inClusterConfig=false
- --sink=influxdb:http://monitoring-influxdb:8086
修改后重新create。重新启动后的heapster pod的日志输出如下:
# kubectl logs -f pod/heapster-hco5i -n kube-system
I0120 02:03:46.014589 1 heapster.go:71] /heapster --source=kubernetes:http://10.47.136.60:8080?inClusterConfig=false --sink=influxdb:http://monitoring-influxdb:8086
I0120 02:03:46.014975 1 heapster.go:72] Heapster version v1.3.0-beta.0
I0120 02:03:46.015080 1 configs.go:60] Using Kubernetes client with master "http://10.47.136.60:8080" and version v1
I0120 02:03:46.015175 1 configs.go:61] Using kubelet port 10255
E0120 02:03:46.025962 1 influxdb.go:217] issues while creating an InfluxDB sink: failed to ping InfluxDB server at "monitoring-influxdb:8086" - Get http://monitoring-influxdb:8086/ping: dial tcp 192.168.3.239:8086: getsockopt: connection refused, will retry on use
I0120 02:03:46.026090 1 influxdb.go:231] created influxdb sink with options: host:monitoring-influxdb:8086 user:root db:k8s
I0120 02:03:46.026214 1 heapster.go:193] Starting with InfluxDB Sink
I0120 02:03:46.026286 1 heapster.go:193] Starting with Metric Sink
I0120 02:03:46.051096 1 heapster.go:105] Starting heapster on port 8082
I0120 02:04:05.211382 1 influxdb.go:209] Created database "k8s" on influxDB server at "monitoring-influxdb:8086"
之前的错误消失了!
我们再次打开Dashboard查看pod信息(这里需要等上一小会儿,因为采集cluster信息也是需要时间的),我们看到集群度量指标信息以图形化的方式展现在我们面前了(可对比本文开头那幅图示):