在诊断Kubernetes集群问题的时候,我们经常注意到集群中某一节点在闪烁*,而这通常是随机的且以奇怪的方式发生。这就是为什么我们一直需要一种工具,它可以测试一个节点与另一个节点之间的可达性,并以Prometheus度量形式呈现结果。有了这个工具,我们还希望在Grafana中创建图表并快速定位发生故障的节点(并在必要时将该节点上所有Pod进行重新调度并进行必要的维护)。
“闪烁”这里我是指某个节点随机变为“NotReady”但之后又恢复正常的某种行为。例如部分流量可能无法到达相邻节点上的Pod。
为什么会发生这种情况?常见原因之一是数据中心交换机中的连接问题。例如,我们曾经在Hetzner中设置一个vswitch,其中一个节点已无法通过该vswitch端口使用,并且恰好在本地网络上完全不可访问。
我们的最后一个要求是可直接在Kubernetes中运行此服务,因此我们将能够通过Helm图表部署所有内容。(例如在使用Ansible的情况下,我们必须为各种环境中的每个角色定义角色:AWS、GCE、裸机等)。由于我们尚未找到针对此环境的现成解决方案,因此我们决定自己来实现。
脚本和配置
我们解决方案的主要组件是一个脚本,该脚本监视每个节点的.status.addresses值。如果某个节点的该值已更改(例如添加了新节点),则我们的脚本使用Helm value方式将节点列表以ConfigMap的形式传递给Helm图表:
- apiVersion: v1
- kind: ConfigMap
- metadata:
- name: ping-exporter-config
- namespace: d8-system
- data:
- nodes.json: >
- {{ .Values.pingExporter.targets | toJson }}
- .Values.pingExporter.targets类似以下:
- "cluster_targets":[{"ipAddress":"192.168.191.11","name":"kube-a-3"},{"ipAddress":"192.168.191.12","name":"kube-a-2"},{"ipAddress":"192.168.191.22","name":"kube-a-1"},{"ipAddress":"192.168.191.23","name":"kube-db-1"},{"ipAddress":"192.168.191.9","name":"kube-db-2"},{"ipAddress":"51.75.130.47","name":"kube-a-4"}],"external_targets":[{"host":"8.8.8.8","name":"google-dns"},{"host":"youtube.com"}]}
下面是Python脚本:
- #!/usr/bin/env python3
- import subprocess
- import prometheus_client
- import re
- import statistics
- import os
- import json
- import glob
- import better_exchook
- import datetime
- better_exchook.install()
- FPING_CMDLINE = "/usr/sbin/fping -p 1000 -C 30 -B 1 -q -r 1".split(" ")
- FPING_REGEX = re.compile(r"^(\S*)\s*: (.*)$", re.MULTILINE)
- CONFIG_PATH = "/config/targets.json"
- registry = prometheus_client.CollectorRegistry()
- prometheus_exceptions_counter = \
- prometheus_client.Counter('kube_node_ping_exceptions', 'Total number of exceptions', [], registry=registry)
- prom_metrics_cluster = {"sent": prometheus_client.Counter('kube_node_ping_packets_sent_total',
- 'ICMP packets sent',
- ['destination_node', 'destination_node_ip_address'],
- registry=registry),
- "received": prometheus_client.Counter('kube_node_ping_packets_received_total',
- 'ICMP packets received',
- ['destination_node', 'destination_node_ip_address'],
- registry=registry),
- "rtt": prometheus_client.Counter('kube_node_ping_rtt_milliseconds_total',
- 'round-trip time',
- ['destination_node', 'destination_node_ip_address'],
- registry=registry),
- "min": prometheus_client.Gauge('kube_node_ping_rtt_min', 'minimum round-trip time',
- ['destination_node', 'destination_node_ip_address'],
- registry=registry),
- "max": prometheus_client.Gauge('kube_node_ping_rtt_max', 'maximum round-trip time',
- ['destination_node', 'destination_node_ip_address'],
- registry=registry),
- "mdev": prometheus_client.Gauge('kube_node_ping_rtt_mdev',
- 'mean deviation of round-trip times',
- ['destination_node', 'destination_node_ip_address'],
- registry=registry)}
- prom_metrics_external = {"sent": prometheus_client.Counter('external_ping_packets_sent_total',
- 'ICMP packets sent',
- ['destination_name', 'destination_host'],
- registry=registry),
- "received": prometheus_client.Counter('external_ping_packets_received_total',
- 'ICMP packets received',
- ['destination_name', 'destination_host'],
- registry=registry),
- "rtt": prometheus_client.Counter('external_ping_rtt_milliseconds_total',
- 'round-trip time',
- ['destination_name', 'destination_host'],
- registry=registry),
- "min": prometheus_client.Gauge('external_ping_rtt_min', 'minimum round-trip time',
- ['destination_name', 'destination_host'],
- registry=registry),
- "max": prometheus_client.Gauge('external_ping_rtt_max', 'maximum round-trip time',
- ['destination_name', 'destination_host'],
- registry=registry),
- "mdev": prometheus_client.Gauge('external_ping_rtt_mdev',
- 'mean deviation of round-trip times',
- ['destination_name', 'destination_host'],
- registry=registry)}
- def validate_envs():
- envs = {"MY_NODE_NAME": os.getenv("MY_NODE_NAME"), "PROMETHEUS_TEXTFILE_DIR": os.getenv("PROMETHEUS_TEXTFILE_DIR"),
- "PROMETHEUS_TEXTFILE_PREFIX": os.getenv("PROMETHEUS_TEXTFILE_PREFIX")}
- for k, v in envs.items():
- if not v:
- raise ValueError("{} environment variable is empty".format(k))
- return envs
- @prometheus_exceptions_counter.count_exceptions()
- def compute_results(results):
- computed = {}
- matches = FPING_REGEX.finditer(results)
- for match in matches:
- host = match.group(1)
- ping_results = match.group(2)
- if "duplicate" in ping_results:
- continue
- splitted = ping_results.split(" ")
- if len(splitted) != 30:
- raise ValueError("ping returned wrong number of results: \"{}\"".format(splitted))
- positive_results = [float(x) for x in splitted if x != "-"]
- if len(positive_results) > 0:
- computed[host] = {"sent": 30, "received": len(positive_results),
- "rtt": sum(positive_results),
- "max": max(positive_results), "min": min(positive_results),
- "mdev": statistics.pstdev(positive_results)}
- else:
- computed[host] = {"sent": 30, "received": len(positive_results), "rtt": 0,
- "max": 0, "min": 0, "mdev": 0}
- if not len(computed):
- raise ValueError("regex match\"{}\" found nothing in fping output \"{}\"".format(FPING_REGEX, results))
- return computed
- @prometheus_exceptions_counter.count_exceptions()
- def call_fping(ips):
- cmdline = FPING_CMDLINE + ips
- process = subprocess.run(cmdline, stdout=subprocess.PIPE,
- stderr=subprocess.STDOUT, universal_newlines=True)
- if process.returncode == 3:
- raise ValueError("invalid arguments: {}".format(cmdline))
- if process.returncode == 4:
- raise OSError("fping reported syscall error: {}".format(process.stderr))
- return process.stdout
- envs = validate_envs()
- files = glob.glob(envs["PROMETHEUS_TEXTFILE_DIR"] + "*")
- for f in files:
- os.remove(f)
- labeled_prom_metrics = {"cluster_targets": [], "external_targets": []}
- while True:
- with open(CONFIG_PATH, "r") as f:
- config = json.loads(f.read())
- config["external_targets"] = [] if config["external_targets"] is None else config["external_targets"]
- for target in config["external_targets"]:
- target["name"] = target["host"] if "name" not in target.keys() else target["name"]
- if labeled_prom_metrics["cluster_targets"]:
- for metric in labeled_prom_metrics["cluster_targets"]:
- if (metric["node_name"], metric["ip"]) not in [(node["name"], node["ipAddress"]) for node in config['cluster_targets']]:
- for k, v in prom_metrics_cluster.items():
- v.remove(metric["node_name"], metric["ip"])
- if labeled_prom_metrics["external_targets"]:
- for metric in labeled_prom_metrics["external_targets"]:
- if (metric["target_name"], metric["host"]) not in [(target["name"], target["host"]) for target in config['external_targets']]:
- for k, v in prom_metrics_external.items():
- v.remove(metric["target_name"], metric["host"])
- labeled_prom_metrics = {"cluster_targets": [], "external_targets": []}
- for node in config["cluster_targets"]:
- metrics = {"node_name": node["name"], "ip": node["ipAddress"], "prom_metrics": {}}
- for k, v in prom_metrics_cluster.items():
- metrics["prom_metrics"][k] = v.labels(node["name"], node["ipAddress"])
- labeled_prom_metrics["cluster_targets"].append(metrics)
- for target in config["external_targets"]:
- metrics = {"target_name": target["name"], "host": target["host"], "prom_metrics": {}}
- for k, v in prom_metrics_external.items():
- metrics["prom_metrics"][k] = v.labels(target["name"], target["host"])
- labeled_prom_metrics["external_targets"].append(metrics)
- out = call_fping([prom_metric["ip"] for prom_metric in labeled_prom_metrics["cluster_targets"]] + \
- [prom_metric["host"] for prom_metric in labeled_prom_metrics["external_targets"]])
- computed = compute_results(out)
- for dimension in labeled_prom_metrics["cluster_targets"]:
- result = computed[dimension["ip"]]
- dimension["prom_metrics"]["sent"].inc(computed[dimension["ip"]]["sent"])
- dimension["prom_metrics"]["received"].inc(computed[dimension["ip"]]["received"])
- dimension["prom_metrics"]["rtt"].inc(computed[dimension["ip"]]["rtt"])
- dimension["prom_metrics"]["min"].set(computed[dimension["ip"]]["min"])
- dimension["prom_metrics"]["max"].set(computed[dimension["ip"]]["max"])
- dimension["prom_metrics"]["mdev"].set(computed[dimension["ip"]]["mdev"])
- for dimension in labeled_prom_metrics["external_targets"]:
- result = computed[dimension["host"]]
- dimension["prom_metrics"]["sent"].inc(computed[dimension["host"]]["sent"])
- dimension["prom_metrics"]["received"].inc(computed[dimension["host"]]["received"])
- dimension["prom_metrics"]["rtt"].inc(computed[dimension["host"]]["rtt"])
- dimension["prom_metrics"]["min"].set(computed[dimension["host"]]["min"])
- dimension["prom_metrics"]["max"].set(computed[dimension["host"]]["max"])
- dimension["prom_metrics"]["mdev"].set(computed[dimension["host"]]["mdev"])
- prometheus_client.write_to_textfile(
envs["PROMETHEUS_TEXTFILE_DIR"] + envs["PROMETHEUS_TEXTFILE_PREFIX"] + envs["MY_NODE_NAME"] + ".prom", registry)
该脚本在每个Kubernetes节点上运行,并且每秒两次发送ICMP数据包到Kubernetes集群的所有实例。收集的结果会存储在文本文件中。
该脚本会包含在Docker镜像中:
- FROM python:3.6-alpine3.8
- COPY rootfs /
- WORKDIR /app
- RUN pip3 install --upgrade pip && pip3 install -r requirements.txt && apk add --no-cache fping
- ENTRYPOINT ["python3", "/app/ping-exporter.py"]
另外,我们还创建了一个ServiceAccount和一个具有唯一权限的对应角色用于获取节点列表(这样我们就可以知道它们的IP地址):
- apiVersion: v1
- kind: ServiceAccount
- metadata:
- name: ping-exporter
- namespace: d8-system
- ---
- kind: ClusterRole
- apiVersion: rbac.authorization.k8s.io/v1
- metadata:
- name: d8-system:ping-exporter
- rules:
- - apiGroups: [""]
- resources: ["nodes"]
- verbs: ["list"]
- ---
- kind: ClusterRoleBinding
- apiVersion: rbac.authorization.k8s.io/v1
- metadata:
- name: d8-system:kube-ping-exporter
- subjects:
- - kind: ServiceAccount
- name: ping-exporter
- namespace: d8-system
- roleRef:
- apiGroup: rbac.authorization.k8s.io
- kind: ClusterRole
- name: d8-system:ping-exporter
最后,我们需要DaemonSet来运行在集群中的所有实例:
- apiVersion: apps/v1
- kind: DaemonSet
- metadata:
- name: ping-exporter
- namespace: d8-system
- spec:
- updateStrategy:
- type: RollingUpdate
- selector:
- matchLabels:
- name: ping-exporter
- template:
- metadata:
- labels:
- name: ping-exporter
- spec:
- terminationGracePeriodSeconds: 0
- tolerations:
- - operator: "Exists"
- hostNetwork: true
- serviceAccountName: ping-exporter
- priorityClassName: cluster-low
- containers:
- - image: private-registry.flant.com/ping-exporter/ping-exporter:v1
- name: ping-exporter
- env:
- - name: MY_NODE_NAME
- valueFrom:
- fieldRef:
- fieldPath: spec.nodeName
- - name: PROMETHEUS_TEXTFILE_DIR
- value: /node-exporter-textfile/
- - name: PROMETHEUS_TEXTFILE_PREFIX
- value: ping-exporter_
- volumeMounts:
- - name: textfile
- mountPath: /node-exporter-textfile
- - name: config
- mountPath: /config
- volumes:
- - name: textfile
- hostPath:
- path: /var/run/node-exporter-textfile
- - name: config
- configMap:
- name: ping-exporter-config
- imagePullSecrets:
- - name: private-registry
该解决方案的最后操作细节是:
- Python脚本执行时,其结果(即存储在主机上/var/run/node-exporter-textfile目录中的文本文件)将传递到DaemonSet类型的node-exporter。
- node-exporter使用--collector.textfile.directory /host/textfile参数启动,这里的/host/textfile是hostPath目录/var/run/node-exporter-textfile。(你可以点击这里了解关于node-exporter中文本文件收集器的更多信息。)
- 最后node-exporter读取这些文件,然后Prometheus从node-exporter实例上收集所有数据。
那么结果如何?
现在该来享受期待已久的结果了。指标创建之后,我们可以使用它们,当然也可以对其进行可视化。以下可以看到它们是怎样的。
首先,有一个通用选择器可让我们在其中选择节点以检查其“源”和“目标”连接。你可以获得一个汇总表,用于在Grafana仪表板中指定的时间段内ping选定节点的结果:
以下是包含有关选定节点的组合统计信息的图形:
另外,我们有一个记录列表,其中每个记录都链接到在“源”节点中选择的每个特定节点的图:
如果将记录展开,你将看到从当前节点到目标节点中已选择的所有其他节点的详细ping统计信息:
下面是相关的图形:
节点之间的ping出现问题的图看起来如何?
如果你在现实生活中观察到类似情况,那就该进行故障排查了!
最后,这是我们对外部主机执行ping操作的可视化效果:
我们可以检查所有节点的总体视图,也可以仅检查任何特定节点的图形:
当你观察到仅影响某些特定节点的连接问题时,这可能会有所帮助。