- 背景 -
最近，在对公司容器云的日志方案进行设计的时候，发现主流的ELK或者EFK比较重，再加上现阶段对于ES复杂的搜索功能很多都用不上最终选择了Grafana开源的Loki日志系统，下面介绍下Loki的背景。
- 缘起 -
当我们的容器云运行的应用或者某个节点出现问题了，解决思路应该如下：
我们的监控使用的是基于Prometheus体系进行改造的，Prometheus中比较重要的是Metric和Alert，Metric是来说明当前或者历史达到了某个值，Alert设置Metric达到某个特定的基数触发了告警，但是这些信息明显是不够的。
我们都知道，Kubernetes的基本单位是Pod，Pod把日志输出到stdout和stderr，平时有什么问题我们通常在界面或者通过命令查看相关的日志，举个例子：当我们的某个Pod的内存变得很大，触发了我们的Alert，这个时候管理员，去页面查询确认是哪个Pod有问题，然后要确认Pod内存变大的原因，我们还需要去查询Pod的日志，如果没有日志系统，那么我们就需要到页面或者使用命令进行查询了：
如果，这个时候应用突然挂了，这个时候我们就无法查到相关的日志了，所以需要引入日志系统，统一收集日志，而使用ELK的话，就需要在Kibana和Grafana之间切换，影响用户体验。所以 ，loki的第一目的就是最小化度量和日志的切换成本，有助于减少异常事件的响应时间和提高用户的体验。
- ELK 存在的问题 -
现有的很多日志采集的方案都是采用全文检索对日志进行索引（如ELK方案），优点是功能丰富，允许复杂的操作。但是，这些方案往往规模复杂，资源占用高，操作苦难。很多功能往往用不上，大多数查询只关注一定时间范围和一些简单的参数（如host、service等），使用这些解决方案就有点杀鸡用牛刀的感觉了。
因此，Loki的第二个目的是，在查询语言的易操作性和复杂性之间可以达到一个权衡。
- 成本 -
全文检索的方案也带来成本问题，简单的说就是全文搜索（如ES）的倒排索引的切分和共享的成本较高。后来出现了其他不同的设计方案如：OKlog，采用最终一致的、基于网格的分布策略。这两个设计决策提供了大量的成本降低和非常简单的操作，但是查询不够方便。因此，Loki的第三个目的是，提高一个更具成本效益的解决方案。
- 架构 -
整体架构
Loki的架构如下：
不难看出，Loki的架构非常简单，使用了和Prometheus一样的标签来作为索引，也就是说，你通过这些标签既可以查询日志的内容也可以查询到监控的数据，不但减少了两种查询之间的切换成本，也极大地降低了日志索引的存储。
Loki将使用与Prometheus相同的服务发现和标签重新标记库，编写了pormtail，在Kubernetes中promtail以DaemonSet方式运行在每个节点中，通过Kubernetes API等到日志的正确元数据，并将它们发送到Loki。下面是日志的存储架构：
- 读写 -
日志数据的写主要依托的是Distributor和Ingester两个组件，整体的流程如下：
- Distributor -
一旦promtail收集日志并将其发送给loki，Distributor就是第一个接收日志的组件。由于日志的写入量可能很大，所以不能在它们传入时将它们写入数据库。这会毁掉数据库。我们需要批处理和压缩数据。
Loki通过构建压缩数据块来实现这一点，方法是在日志进入时对其进行gzip操作，组件ingester是一个有状态的组件，负责构建和刷新chunck，当chunk达到一定的数量或者时间后，刷新到存储中去。每个流的日志对应一个ingester，当日志到达Distributor后，根据元数据和hash算法计算出应该到哪个ingester上面。
此外，为了冗余和弹性，我们将其复制n（默认情况下为3）次。
- Ingester -
Ingester接收到日志并开始构建chunk：
基本上就是将日志进行压缩并附加到chunk上面。一旦chunk“填满”（数据达到一定数量或者过了一定期限），ingester将其刷新到数据库。我们对块和索引使用单独的数据库，因为它们存储的数据类型不同。
刷新一个chunk之后，ingester然后创建一个新的空chunk并将新条目添加到该chunk中。
- Querier -
读取就非常简单了，由Querier负责给定一个时间范围和标签选择器，Querier查看索引以确定哪些块匹配，并通过greps将结果显示出来。它还从Ingester获取尚未刷新的最新数据。
对于每个查询，一个查询器将为您显示所有相关日志。实现了查询并行化，提供分布式grep，使即使是大型查询也是足够的。
- 可扩展性 -
Loki的索引存储可以是cassandra/bigtable/dynamodb，而chuncks可以是各种对象存储，Querier和Distributor都是无状态的组件。对于ingester他虽然是有状态的但是，当新的节点加入或者减少，整节点间的chunk会重新分配，已适应新的散列环。而Loki底层存储的实现Cortex已经 在实际的生产中投入使用多年了。有了这句话，我可以放心的在环境中实验一把了。
- 部署 -
Loki的安装非常简单。
创建namespace
oc new-project loki
权限设置
oc adm policy add-scc-to-user anyuid -z default -n loki
oc adm policy add-cluster-role-to-user cluster-admin system:serviceaccount:loki:default
安装Loki
安装命令：
oc create -f statefulset.json -n loki
statefulset.json如下：
{
"apiVersion": "apps/v1",
"kind": "StatefulSet",
"metadata": {
"name": "loki"
},
"spec": {
"podManagementPolicy": "OrderedReady",
"replicas": 1,
"revisionHistoryLimit": 10,
"selector": {
"matchLabels": {
"app": "loki"
}
},
"serviceName": "womping-stoat-loki-headless",
"template": {
"metadata": {
"annotations": {
"checksum/config": "da297d66ee53e0ce68b58e12be7ec5df4a91538c0b476cfe0ed79666343df72b",
"prometheus.io/port": "http-metrics",
"prometheus.io/scrape": "true"
},
"creationTimestamp": null,
"labels": {
"app": "loki",
"name": "loki"
}
},
"spec": {
"affinity": {},
"containers": [
{
"args": [
"-config.file=/etc/loki/local-config.yaml"
],
"image": "grafana/loki:latest",
"imagePullPolicy": "IfNotPresent",
"livenessProbe": {
"failureThreshold": 3,
"httpGet": {
"path": "/ready",
"port": "http-metrics",
"scheme": "HTTP"
},
"initialDelaySeconds": 45,
"periodSeconds": 10,
"successThreshold": 1,
"timeoutSeconds": 1
},
"name": "loki",
"ports": [
{
"containerPort": 3100,
"name": "http-metrics",
"protocol": "TCP"
}
],
"readinessProbe": {
"failureThreshold": 3,
"httpGet": {
"path": "/ready",
"port": "http-metrics",
"scheme": "HTTP"
},
"initialDelaySeconds": 45,
"periodSeconds": 10,
"successThreshold": 1,
"timeoutSeconds": 1
},
"resources": {},
"terminationMessagePath": "/dev/termination-log",
"terminationMessagePolicy": "File",
"volumeMounts": [
{
"mountPath": "/tmp/loki",
"name": "storage"
}
]
}
],
"dnsPolicy": "ClusterFirst",
"restartPolicy": "Always",
"schedulerName": "default-scheduler",
"terminationGracePeriodSeconds": 30,
"volumes": [
{
"emptyDir": {},
"name": "storage"
}
]
}
},
"updateStrategy": {
"type": "RollingUpdate"
}
}
}
安装Promtail
安装命令：
oc create -f configmap.json -n loki
configmap.json如下：
{
"apiVersion": "v1",
"data": {
/*
* 提示：该行代码过长，系统自动注释不进行高亮。一键复制会移除系统注释
* "promtail.yaml": "client:\n backoff_config:\n maxbackoff: 5s\n maxretries: 5\n minbackoff: 100ms\n batchsize: 102400\n batchwait: 1s\n external_labels: {}\n timeout: 10s\npositions:\n filename: /run/promtail/positions.yaml\nserver:\n http_listen_port: 3101\ntarget_config:\n sync_period: 10s\n\nscrape_configs:\n- job_name: kubernetes-pods-name\n pipeline_stages:\n - docker: {}\n \n kubernetes_sd_configs:\n - role: pod\n relabel_configs:\n - source_labels:\n - __meta_kubernetes_pod_label_name\n target_label: __service__\n - source_labels:\n - __meta_kubernetes_pod_node_name\n target_label: __host__\n - action: drop\n regex: ^$\n source_labels:\n - __service__\n - action: labelmap\n regex: __meta_kubernetes_pod_label_(.+)\n - action: replace\n replacement: $1\n separator: /\n source_labels:\n - __meta_kubernetes_namespace\n - __service__\n target_label: job\n - action: replace\n source_labels:\n - __meta_kubernetes_namespace\n target_label: namespace\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_name\n target_label: instance\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_container_name\n target_label: container_name\n - replacement: /var/log/pods/*$1/*.log\n separator: /\n source_labels:\n - __meta_kubernetes_pod_uid\n - __meta_kubernetes_pod_container_name\n target_label: __path__\n- job_name: kubernetes-pods-app\n pipeline_stages:\n - docker: {}\n \n kubernetes_sd_configs:\n - role: pod\n relabel_configs:\n - action: drop\n regex: .+\n source_labels:\n - __meta_kubernetes_pod_label_name\n - source_labels:\n - __meta_kubernetes_pod_label_app\n target_label: __service__\n - source_labels:\n - __meta_kubernetes_pod_node_name\n target_label: __host__\n - action: drop\n regex: ^$\n source_labels:\n - __service__\n - action: labelmap\n regex: __meta_kubernetes_pod_label_(.+)\n - action: replace\n replacement: $1\n separator: /\n source_labels:\n - __meta_kubernetes_namespace\n - __service__\n target_label: job\n - action: replace\n source_labels:\n - __meta_kubernetes_namespace\n target_label: namespace\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_name\n target_label: instance\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_container_name\n target_label: container_name\n - replacement: /var/log/pods/*$1/*.log\n separator: /\n source_labels:\n - __meta_kubernetes_pod_uid\n - __meta_kubernetes_pod_container_name\n target_label: __path__\n- job_name: kubernetes-pods-direct-controllers\n pipeline_stages:\n - docker: {}\n \n kubernetes_sd_configs:\n - role: pod\n relabel_configs:\n - action: drop\n regex: .+\n separator: ''\n source_labels:\n - __meta_kubernetes_pod_label_name\n - __meta_kubernetes_pod_label_app\n - action: drop\n regex: ^([0-9a-z-.]+)(-[0-9a-f]{8,10})$\n source_labels:\n - __meta_kubernetes_pod_controller_name\n - source_labels:\n - __meta_kubernetes_pod_controller_name\n target_label: __service__\n - source_labels:\n - __meta_kubernetes_pod_node_name\n target_label: __host__\n - action: drop\n regex: ^$\n source_labels:\n - __service__\n - action: labelmap\n regex: __meta_kubernetes_pod_label_(.+)\n - action: replace\n replacement: $1\n separator: /\n source_labels:\n - __meta_kubernetes_namespace\n - __service__\n target_label: job\n - action: replace\n source_labels:\n - __meta_kubernetes_namespace\n target_label: namespace\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_name\n target_label: instance\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_container_name\n target_label: container_name\n - replacement: /var/log/pods/*$1/*.log\n separator: /\n source_labels:\n - __meta_kubernetes_pod_uid\n - __meta_kubernetes_pod_container_name\n target_label: __path__\n- job_name: kubernetes-pods-indirect-controller\n pipeline_stages:\n - docker: {}\n \n kubernetes_sd_configs:\n - role: pod\n relabel_configs:\n - action: drop\n regex: .+\n separator: ''\n source_labels:\n - __meta_kubernetes_pod_label_name\n - __meta_kubernetes_pod_label_app\n - action: keep\n regex: ^([0-9a-z-.]+)(-[0-9a-f]{8,10})$\n source_labels:\n - __meta_kubernetes_pod_controller_name\n - action: replace\n regex: ^([0-9a-z-.]+)(-[0-9a-f]{8,10})$\n source_labels:\n - __meta_kubernetes_pod_controller_name\n target_label: __service__\n - source_labels:\n - __meta_kubernetes_pod_node_name\n target_label: __host__\n - action: drop\n regex: ^$\n source_labels:\n - __service__\n - action: labelmap\n regex: __meta_kubernetes_pod_label_(.+)\n - action: replace\n replacement: $1\n separator: /\n source_labels:\n - __meta_kubernetes_namespace\n - __service__\n target_label: job\n - action: replace\n source_labels:\n - __meta_kubernetes_namespace\n target_label: namespace\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_name\n target_label: instance\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_container_name\n target_label: container_name\n - replacement: /var/log/pods/*$1/*.log\n separator: /\n source_labels:\n - __meta_kubernetes_pod_uid\n - __meta_kubernetes_pod_container_name\n target_label: __path__\n- job_name: kubernetes-pods-static\n pipeline_stages:\n - docker: {}\n \n kubernetes_sd_configs:\n - role: pod\n relabel_configs:\n - action: drop\n regex: ^$\n source_labels:\n - __meta_kubernetes_pod_annotation_kubernetes_io_config_mirror\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_label_component\n target_label: __service__\n - source_labels:\n - __meta_kubernetes_pod_node_name\n target_label: __host__\n - action: drop\n regex: ^$\n source_labels:\n - __service__\n - action: labelmap\n regex: __meta_kubernetes_pod_label_(.+)\n - action: replace\n replacement: $1\n separator: /\n source_labels:\n - __meta_kubernetes_namespace\n - __service__\n target_label: job\n - action: replace\n source_labels:\n - __meta_kubernetes_namespace\n target_label: namespace\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_name\n target_label: instance\n - action: replace\n source_labels:\n - __meta_kubernetes_pod_container_name\n target_label: container_name\n - replacement: /var/log/pods/*$1/*.log\n separator: /\n source_labels:\n - __meta_kubernetes_pod_annotation_kubernetes_io_config_mirror\n - __meta_kubernetes_pod_container_name\n target_label: __path__\n"
*/
},
"kind": "ConfigMap",
"metadata": {
"creationTimestamp": "2019-09-05T01:05:03Z",
"labels": {
"app": "promtail",
"chart": "promtail-0.12.0",
"heritage": "Tiller",
"release": "lame-zorse"
},
"name": "lame-zorse-promtail",
"namespace": "loki",
"resourceVersion": "17921611",
"selfLink": "/api/v1/namespaces/loki/configmaps/lame-zorse-promtail",
"uid": "30fcb896-cf79-11e9-b58e-e4a8b6cc47d2"
}
}
oc create -f daemonset.json -n loki
daemonset.json如下：
{
"apiVersion": "apps/v1",
"kind": "DaemonSet",
"metadata": {
"annotations": {
"deployment.kubernetes.io/revision": "2"
},
"creationTimestamp": "2019-09-05T01:16:37Z",
"generation": 2,
"labels": {
"app": "promtail",
"chart": "promtail-0.12.0",
"heritage": "Tiller",
"release": "lame-zorse"
},
"name": "lame-zorse-promtail",
"namespace": "loki"
},
"spec": {
"progressDeadlineSeconds": 600,
"replicas": 1,
"revisionHistoryLimit": 10,
"selector": {
"matchLabels": {
"app": "promtail",
"release": "lame-zorse"
}
},
"strategy": {
"rollingUpdate": {
"maxSurge": 1,
"maxUnavailable": 1
},
"type": "RollingUpdate"
},
"template": {
"metadata": {
"annotations": {
"checksum/config": "75a25ee4f2869f54d394bf879549a9c89c343981a648f8d878f69bad65dba809",
"prometheus.io/port": "http-metrics",
"prometheus.io/scrape": "true"
},
"creationTimestamp": null,
"labels": {
"app": "promtail",
"release": "lame-zorse"
}
},
"spec": {
"affinity": {},
"containers": [
{
"args": [
"-config.file=/etc/promtail/promtail.yaml",
"-client.url=http://loki.loki.svc:3100/api/prom/push"
],
"env": [
{
"name": "HOSTNAME",
"valueFrom": {
"fieldRef": {
"apiVersion": "v1",
"fieldPath": "spec.nodeName"
}
}
}
],
"image": "grafana/promtail:v0.3.0",
"imagePullPolicy": "IfNotPresent",
"name": "promtail",
"ports": [
{
"containerPort": 3101,
"name": "http-metrics",
"protocol": "TCP"
}
],
"readinessProbe": {
"failureThreshold": 5,
"httpGet": {
"path": "/ready",
"port": "http-metrics",
"scheme": "HTTP"
},
"initialDelaySeconds": 10,
"periodSeconds": 10,
"successThreshold": 1,
"timeoutSeconds": 1
},
"resources": {},
"securityContext": {
"readOnlyRootFilesystem": true,
"runAsUser": 0
},
"terminationMessagePath": "/dev/termination-log",
"terminationMessagePolicy": "File",
"volumeMounts": [
{
"mountPath": "/etc/promtail",
"name": "config"
},
{
"mountPath": "/run/promtail",
"name": "run"
},
{
"mountPath": "/var/lib/docker/containers",
"name": "docker",
"readOnly": true
},
{
"mountPath": "/var/log/pods",
"name": "pods",
"readOnly": true
}
]
}
],
"dnsPolicy": "ClusterFirst",
"restartPolicy": "Always",
"schedulerName": "default-scheduler",
"securityContext": {},
"terminationGracePeriodSeconds": 30,
"volumes": [
{
"configMap": {
"defaultMode": 420,
"name": "lame-zorse-promtail"
},
"name": "config"
},
{
"hostPath": {
"path": "/run/promtail",
"type": ""
},
"name": "run"
},
{
"hostPath": {
"path": "/var/lib/docker/containers",
"type": ""
},
"name": "docker"
},
{
"hostPath": {
"path": "/var/log/pods",
"type": ""
},
"name": "pods"
}
]
}
}
}
}
安装服务
oc create -f service.json -n loki
service.json的内容如下：
{
"apiVersion": "v1",
"kind": "Service",
"metadata": {
"creationTimestamp": "2019-09-04T09:37:49Z",
"name": "loki",
"namespace": "loki",
"resourceVersion": "17800188",
"selfLink": "/api/v1/namespaces/loki/services/loki",
"uid": "a87fe237-cef7-11e9-b58e-e4a8b6cc47d2"
},
"spec": {
"externalTrafficPolicy": "Cluster",
"ports": [
{
"name": "lokiport",
"port": 3100,
"protocol": "TCP",
"targetPort": 3100
}
],
"selector": {
"app": "loki"
},
"sessionAffinity": "None",
"type": "NodePort"
},
"status": {
"loadBalancer": {}
}
- 语法 -
Loki提供了HTTP接口，我们这里就不详解了，大家可以看：https://github.com/grafana/loki/blob/master/docs/api.md
我们这里说下查询的接口如何使用。
第一步，获取当前Loki的元数据类型：
curl http://192.168.25.30:30972/api/prom/label
{
"values": ["alertmanager", "app", "component", "container_name", "controller_revision_hash", "deployment", "deploymentconfig", "docker_registry", "draft", "filename", "instance", "job", "logging_infra", "metrics_infra", "name", "namespace", "openshift_io_component", "pod_template_generation", "pod_template_hash", "project", "projectname", "prometheus", "provider", "release", "router", "servicename", "statefulset_kubernetes_io_pod_name", "stream", "tekton_dev_pipeline", "tekton_dev_pipelineRun", "tekton_dev_pipelineTask", "tekton_dev_task", "tekton_dev_taskRun", "type", "webconsole"]
}
第二步，获取某个元数据类型的值：
curl http://192.168.25.30:30972/api/prom/label/namespace/values
{"values":["cicd","default","gitlab","grafanaserver","jenkins","jx-staging","kube-system","loki","mysql-exporter","new2","openshift-console","openshift-infra","openshift-logging","openshift-monitoring","openshift-node","openshift-sdn","openshift-web-console","tekton-pipelines","test111"]}
第三步，根据label进行查询，例如：
http://192.168.25.30:30972/api/prom/query?direction=BACKWARD&limit=1000®exp=&query={namespace="cicd"}&start=1567644457221000000&end=1567730857221000000&refId=A
参数解析：
- query：一种查询语法详细见下面章节，{name=~“mysql.+”} or {namespace=“cicd”} |= "error"表示查询，namespace为CI/CD的日志中，有error字样的信息。
- limit：返回日志的数量
- start：开始时间，Unix时间表示方法 默认为，一小时前时间
- end：结束时间，默认为当前时间
- direction：forward或者backward，指定limit时候有用，默认为 backward
- regexp：对结果进行regex过滤
- LogQL 语法 -
选择器
对于查询表达式的标签部分，将放在{}中，多个标签表达式用逗号分隔：
{app="mysql",name="mysql-backup"}
支持的符号有：
- =：完全相同。
- !=：不平等。
- =~：正则表达式匹配。
- !~：不要正则表达式匹配。
- 过滤表达式 -
编写日志流选择器后，您可以通过编写搜索表达式进一步过滤结果。搜索表达式可以文本或正则表达式。
如：
- {job=“mysql”} |= “error”
- {name=“kafka”} |~ “tsdb-ops.*io:2003”
- {instance=~“kafka-[23]”,name=“kafka”} != kafka.server:type=ReplicaManager
支持多个过滤：
- {job=“mysql”} |= “error” != “timeout”
目前支持的操作符：
- |= line包含字符串。
- != line不包含字符串。
- |~ line匹配正则表达式。
- !~ line与正则表达式不匹配。
表达式遵循https://github.com/google/re2/wiki/Syntax语法。