EFK Stack on Kubernetes (Part 2)

This is the final part of our Kubernetes logging series. In case you missed part 1 you can find it here. In this tutorial, we will learn about configuring Filebeat to run as a DaemonSet in our Kubernetes cluster in order to ship logs to the Elasticsearch backend. We are using Filebeat instead of FluentD or FluentBit because it is an extremely lightweight utility and has a first-class support for Kubernetes. It is best for production-level setups.

1. Deployment Architecture

Filebeat will run as a DaemonSet in our Kubernetes cluster. It will be:

• Deployed in a separate namespace called Logging.
• Pods will be scheduled on both Master nodes and Worker Nodes.
• Master Node pods will forward api-server logs for audit and cluster administration purposes.
• Client Node pods will forward workload related logs for application observability.

2.1 Creating Filebeat ServiceAccount and ClusterRole

Deploy the following manifest to create the required permissions for Filebeat pods.

apiVersion: v1kind: Namespacemetadata:  name: logging---apiVersion: v1kind: ServiceAccountmetadata:  name: filebeat  namespace: logging  labels:    k8s-app: filebeat---apiVersion: rbac.authorization.k8s.io/v1beta1kind: ClusterRolemetadata:  name: filebeat  namespace: logging  labels:    k8s-app: filebeatrules:- apiGroups: [""] # "" indicates the core API group  resources:  - namespaces  - pods  verbs:  - get  - watch  - list---apiVersion: rbac.authorization.k8s.io/v1beta1kind: ClusterRoleBindingmetadata:  name: filebeat  namespace: loggingsubjects:- kind: ServiceAccount  name: filebeat  namespace: kube-systemroleRef:  kind: ClusterRole  name: filebeat  apiGroup: rbac.authorization.k8s.io

We should make sure that ClusterRole permissions are as limited as possible from a security point of view. If either of the pod associated with this service account gets compromised then the attacker would not be able to gain access entire cluster or applications running in it.

2.2 Creating Filebeat ConfigMap

Use the following manifest to create a ConfigMap which will be used by Filebeat pods.

apiVersion: v1kind: Namespacemetadata:  name: logging---apiVersion: v1kind: ConfigMapmetadata:  name: filebeat-config  namespace: logging  labels:    k8s-app: filebeat    kubernetes.io/cluster-service: "true"data:  filebeat.yml: |-    filebeat.config:    #  inputs:    #    path: ${path.config}/inputs.d/*.yml    #    reload.enabled: true      modules:        path: ${path.config}/modules.d/*.yml        reload.enabled: true    filebeat.autodiscover:      providers:        - type: kubernetes          hints.enabled: true          include_annotations: ["artifact.spinnaker.io/name","ad.datadoghq.com/tags"]          include_labels: ["app.kubernetes.io/name"]          labels.dedot: true          annotations.dedot: true          templates:            - condition:                equals:                  kubernetes.namespace: myapp   #Set the namespace in which your app is running, can add multiple conditions in case of more than 1 namespace.              config:                - type: docker                  containers.ids:                    - "${data.kubernetes.container.id}"                  multiline:                    pattern: '^[A-Za-z ]+[0-9]{2} (?:[01]\d|2[0123]):(?:[012345]\d):(?:[012345]\d)'.   #Timestamp regex for the app logs. Change it as per format.                     negate: true                    match: after            - condition:                equals:                  kubernetes.namespace: elasticsearch              config:                - type: docker                  containers.ids:                    - "${data.kubernetes.container.id}"                  multiline:                    pattern: '^\[[0-9]{4}-[0-9]{2}-[0-9]{2}|^[0-9]{4}-[0-9]{2}-[0-9]{2}T'                    negate: true                    match: after                        processors:      - add_cloud_metadata: ~      - drop_fields:          when:            has_fields: ['kubernetes.labels.app']          fields:            - 'kubernetes.labels.app'    output.elasticsearch:      hosts: ['${ELASTICSEARCH_HOST:elasticsearch}:${ELASTICSEARCH_PORT:9200}']

Important concepts for the Filebeat ConfigMap:

• hints.enabled: This activates Filebeat’s hints module for Kubernetes. By using this we can use pod annotations to pass config directly to Filebeat pod. We can specify different multiline patterns and various other types of config. More about this can be read here.
• include_annotations: Setting this to true enables Filebeat to retain any pod annotation for a particular log entry. These annotations can be later used to filter logs in the Kibana console.
• include_labels: Setting this to true enables Filebeat to retain any pod labels for a particular log entry. These labels can be later used to filter logs in the Kibana console.
• We can also filter logs for a particular namespace and then can process the log entries accordingly. Here docker log processor is used. We can also use different multiline patterns for different namespaces.
• The output is set to Elasticsearch because we are using Elasticsearch as the storage backend. Alternatively, this can also point to Redis, Logstash, Kafka or even a File. More this can be read here.
• Cloud metadata processor includes some host-specific fields in the log entry. This is helpful when we try to filter logs specific to a particular worker node.

2.3 Deploying Filebeat DaemonSet

Use the manifest below to deploy the Filebeat DaemonSet.

apiVersion: v1kind: Namespacemetadata:  name: logging---apiVersion: extensions/v1beta1kind: DaemonSetmetadata:  name: filebeat  namespace: logging  labels:    k8s-app: filebeatspec:  template:    metadata:      labels:        k8s-app: filebeat    spec:      serviceAccountName: filebeat      terminationGracePeriodSeconds: 30      tolerations:      - effect: NoSchedule        key: node-role.kubernetes.io/master      containers:      - name: filebeat        image: elastic/filebeat:6.5.4        args: [          "-c", "/usr/share/filebeat/filebeat.yml",          "-e",        ]        env:        - name: ELASTICSEARCH_HOST          value: elasticsearch.elasticsearch        - name: ELASTICSEARCH_PORT          value: "9200"        securityContext:          runAsUser: 0          # If using Red Hat OpenShift uncomment this:          #privileged: true        resources:          limits:            memory: 200Mi          requests:            cpu: 100m            memory: 100Mi        volumeMounts:        - name: config          mountPath: /usr/share/filebeat/filebeat.yml          readOnly: true          subPath: filebeat.yml        - name: inputs          mountPath: /usr/share/filebeat/inputs.d          readOnly: true        - name: data          mountPath: /usr/share/filebeat/data        - name: varlibdockercontainers          mountPath: /var/lib/docker/containers          readOnly: true      volumes:      - name: config        configMap:          defaultMode: 0600          name: filebeat-config      - name: varlibdockercontainers        hostPath:          path: /var/lib/docker/containers      - name: inputs        configMap:          defaultMode: 0600          name: filebeat-inputs      # data folder stores a registry of read status for all files, so we don't send everything again on a Filebeat pod restart      - name: data        hostPath:          path: /var/lib/filebeat-data          type: DirectoryOrCreate---

Let’s see what is going on here:

• Logs for each pod are written to /var/log/docker/containers. We are mounting this directory from the host to the Filebeat pod and then Filebeat processes the logs according to the provided configuration.
• We have set the env var ELASTICSEARCH_HOST to elasticsearch.elasticsearch to refer to the Elasticsearch client service which was created in part 1 of this article. In case you already have an Elasticsearch cluster running, the env var should be set to point to it.

Please note the following setting in the manifest:

...      tolerations:      - effect: NoSchedule        key: node-role.kubernetes.io/master...

This makes sure that our Filebeat DaemonSet schedules a pod on the master node as well.
Once the Filebeat DaemonSet is deployed we can check if our pods get scheduled properly.

root$ kubectl -n logging get pods  -o wideNAME             READY   STATUS    RESTARTS   AGE   IP            NODE                                         NOMINATED NODE   READINESS GATESfilebeat-4kchs   1/1     Running   0          6d    100.96.8.2    ip-10-10-30-206.us-east-2.compute.internal   <none>           <none>filebeat-6nrpc   1/1     Running   0          6d    100.96.7.6    ip-10-10-29-252.us-east-2.compute.internal   <none>           <none>filebeat-7qs2s   1/1     Running   0          6d    100.96.1.6    ip-10-10-30-161.us-east-2.compute.internal   <none>           <none>filebeat-j5xz6   1/1     Running   0          6d    100.96.5.3    ip-10-10-28-186.us-east-2.compute.internal   <none>           <none>filebeat-pskg5   1/1     Running   0          6d    100.96.64.4   ip-10-10-29-142.us-east-2.compute.internal   <none>           <none>filebeat-vjdtg   1/1     Running   0          6d    100.96.65.3   ip-10-10-30-118.us-east-2.compute.internal   <none>           <none>filebeat-wm24j   1/1     Running   0          6d    100.96.0.4    ip-10-10-28-162.us-east-2.compute.internal   <none>           <none>root$ kubectl -get nodes -o wideNAME                                         STATUS   ROLES    AGE   VERSION   INTERNAL-IP    EXTERNAL-IP   OS-IMAGE                       KERNEL-VERSION   CONTAINER-RUNTIMEip-10-10-28-162.us-east-2.compute.internal   Ready    master   6d    v1.14.8   10.10.28.162   <none>        Debian GNU/Linux 9 (stretch)   4.9.0-9-amd64    docker://18.6.3ip-10-10-28-186.us-east-2.compute.internal   Ready    node     6d    v1.14.8   10.10.28.186   <none>        Debian GNU/Linux 9 (stretch)   4.9.0-9-amd64    docker://18.6.3ip-10-10-29-142.us-east-2.compute.internal   Ready    master   6d    v1.14.8   10.10.29.142   <none>        Debian GNU/Linux 9 (stretch)   4.9.0-9-amd64    docker://18.6.3ip-10-10-29-252.us-east-2.compute.internal   Ready    node     6d    v1.14.8   10.10.29.252   <none>        Debian GNU/Linux 9 (stretch)   4.9.0-9-amd64    docker://18.6.3ip-10-10-30-118.us-east-2.compute.internal   Ready    master   6d    v1.14.8   10.10.30.118   <none>        Debian GNU/Linux 9 (stretch)   4.9.0-9-amd64    docker://18.6.3ip-10-10-30-161.us-east-2.compute.internal   Ready    node     6d    v1.14.8   10.10.30.161   <none>        Debian GNU/Linux 9 (stretch)   4.9.0-9-amd64    docker://18.6.3ip-10-10-30-206.us-east-2.compute.internal   Ready    node     6d    v1.14.8   10.10.30.206   <none>        Debian GNU/Linux 9 (stretch)   4.9.0-9-amd64    docker://18.6.3

If we tail the logs for one of the pods we can clearly see that it connected to Elasticsearch and has started harvester for the files. The snippet below shows this:

2019-11-19T06:22:03.435Z	INFO	log/input.go:138	Configured paths: [/var/lib/docker/containers/c2b29f5e06eb8affb2cce7cf2501f6f824a2fd83418d09823faf4e74a5a51eb7/*.log]2019-11-19T06:22:03.435Z	INFO	input/input.go:114	Starting input of type: docker; ID: 4134444498769889169 2019-11-19T06:22:04.786Z	INFO	input/input.go:149	input ticker stopped2019-11-19T06:22:04.786Z	INFO	input/input.go:167	Stopping Input: 41344444987698891692019-11-19T06:22:19.295Z	INFO	[monitoring]	log/log.go:144	Non-zero metrics in the last 30s	{"monitoring": {"metrics": {"beat":{"cpu":{"system":{"ticks":641680,"time":{"ms":16}},"total":{"ticks":2471920,"time":{"ms":180},"value":2471920},"user":{"ticks":1830240,"time":{"ms":164}}},"handles":{"limit":{"hard":1048576,"soft":1048576},"open":20},"info":{"ephemeral_id":"007e8090-7c62-4b44-97fb-e74e8177dc54","uptime":{"ms":549390018}},"memstats":{"gc_next":47281968,"memory_alloc":29021760,"memory_total":156062982472}},"filebeat":{"events":{"added":111,"done":111},"harvester":{"closed":2,"open_files":15,"running":13}},"libbeat":{"config":{"module":{"running":0}},"output":{"events":{"acked":108,"batches":15,"total":108},"read":{"bytes":69},"write":{"bytes":123536}},"pipeline":{"clients":1847,"events":{"active":0,"filtered":3,"published":108,"total":111},"queue":{"acked":108}}},"registrar":{"states":{"current":87,"update":111},"writes":{"success":18,"total":18}},"system":{"load":{"1":0.98,"15":1.71,"5":1.59,"norm":{"1":0.0613,"15":0.1069,"5":0.0994}}}}}}2019-11-19T06:22:49.295Z	INFO	[monitoring]	log/log.go:144	Non-zero metrics in the last 30s	{"monitoring": {"metrics": {"beat":{"cpu":{"system":{"ticks":641720,"time":{"ms":44}},"total":{"ticks":2472030,"time":{"ms":116},"value":2472030},"user":{"ticks":1830310,"time":{"ms":72}}},"handles":{"limit":{"hard":1048576,"soft":1048576},"open":20},"info":{"ephemeral_id":"007e8090-7c62-4b44-97fb-e74e8177dc54","uptime":{"ms":549420018}},"memstats":{"gc_next":47281968,"memory_alloc":38715472,"memory_total":156072676184}},"filebeat":{"events":{"active":12,"added":218,"done":206},"harvester":{"open_files":15,"running":13}},"libbeat":{"config":{"module":{"running":0}},"output":{"events":{"acked":206,"batches":24,"total":206},"read":{"bytes":102},"write":{"bytes":269666}},"pipeline":{"clients":1847,"events":{"active":12,"published":218,"total":218},"queue":{"acked":206}}},"registrar":{"states":{"current":87,"update":206},"writes":{"success":24,"total":24}},"system":{"load":{"1":1.22,"15":1.7,"5":1.58,"norm":{"1":0.0763,"15":0.1063,"5":0.0988}}}}}}2019-11-19T06:23:19.295Z	INFO	[monitoring]	log/log.go:144	Non-zero metrics in the last 30s	{"monitoring": {"metrics": {"beat":{"cpu":{"system":{"ticks":641750,"time":{"ms":28}},"total":{"ticks":2472110,"time":{"ms":72},"value":2472110},"user":{"ticks":1830360,"time":{"ms":44}}},"handles":{"limit":{"hard":1048576,"soft":1048576},"open":20},"info":{"ephemeral_id":"007e8090-7c62-4b44-97fb-e74e8177dc54","uptime":{"ms":549450017}},"memstats":{"gc_next":47281968,"memory_alloc":43140256,"memory_total":156077100968}},"filebeat":{"events":{"active":-12,"added":43,"done":55},"harvester":{"open_files":15,"running":13}},"libbeat":{"config":{"module":{"running":0}},"output":{"events":{"acked":55,"batches":12,"total":55},"read":{"bytes":51},"write":{"bytes":70798}},"pipeline":{"clients":1847,"events":{"active":0,"published":43,"total":43},"queue":{"acked":55}}},"registrar":{"states":{"current":87,"update":55},"writes":{"success":12,"total":12}},"system":{"load":{"1":0.99,"15":1.67,"5":1.49,"norm":{"1":0.0619,"15":0.1044,"5":0.0931}}}}}}2019-11-19T06:23:25.261Z	INFO	log/harvester.go:255	Harvester started for file: /var/lib/docker/containers/ccb7dc75ecc755734f6befc4965b9fdae74d59810914101eadf63daa69eb62e2/ccb7dc75ecc755734f6befc4965b9fdae74d59810914101eadf63daa69eb62e2-json.log2019-11-19T06:23:49.295Z	INFO	[monitoring]	log/log.go:144	Non-zero metrics in the last 30s	{"monitoring": {"metrics": {"beat":{"cpu":{"system":{"ticks":641780,"time":{"ms":28}},"total":{"ticks":2472310,"time":{"ms":196},"value":2472310},"user":{"ticks":1830530,"time":{"ms":168}}},"handles":{"limit":{"hard":1048576,"soft":1048576},"open":21},"info":{"ephemeral_id":"007e8090-7c62-4b44-97fb-e74e8177dc54","uptime":{"ms":549480018}},"memstats":{"gc_next":47789200,"memory_alloc":31372376,"memory_total":156086697176,"rss":-1064960}},"filebeat":{"events":{"active":16,"added":170,"done":154},"harvester":{"open_files":16,"running":14,"started":1}},"libbeat":{"config":{"module":{"running":0}},"output":{"events":{"acked":153,"batches":24,"total":153},"read":{"bytes":115},"write":{"bytes":207569}},"pipeline":{"clients":1847,"events":{"active":16,"filtered":1,"published":169,"total":170},"queue":{"acked":153}}},"registrar":{"states":{"current":87,"update":154},"writes":{"success":25,"total":25}},"system":{"load":{"1":0.87,"15":1.63,"5":1.41,"norm":{"1":0.0544,"15":0.1019,"5":0.0881}}}}}}

Once we have all our pods running, then we can create an index pattern of the type filebeat-* in Kibana. Filebeat indexes are generally timestamped. As soon as we create the index pattern, all the searchable available fields can be seen and should be imported.

Lastly, we can search through our application logs and create dashboards if needed.
It is highly recommended to have JSON logger in our applications because it makes log processing extremely easy and messages can be parsed easily.

Conclusion

This concludes our logging set-up. All of the provided configuration files have been tested in production environments and are readily deployable.

Discover more with Gcore Managed Kubernetes