k8s

Kubernetes plugin

It reads Kubernetes logs and also adds pod meta-information. Also, it joins split logs into a single event.

We recommend using the Helm-chart for running in Kubernetes

Source log file should be named in the following format:
[pod-name]_[namespace]_[container-name]-[container-id].log

E.g. my_pod-1566485760-trtrq_my-namespace_my-container-4e0301b633eaa2bfdcafdeba59ba0c72a3815911a6a820bf273534b0f32d98e0.log

An information which plugin adds:

• k8s_node – node name where pod is running;
• k8s_node_label_* – node labels;
• k8s_pod – pod name;
• k8s_namespace – pod namespace name;
• k8s_container – pod container name;
• k8s_label_* – pod labels.

⚠ Use add_file_name plugin if you want to add filename to events.

Example:

pipelines:
  example_k8s_pipeline:
    input:
      type: k8s
      offsets_file: /data/offsets.yaml
      file_config:                        // customize file plugin
        persistence_mode: sync
        read_buffer_size: 2048

To allow the plugin to access the necessary Kubernetes resources, you need to create a ClusterRole that grants permissions to read pod and node information.

apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: filed-pod-watcher
rules:
- apiGroups: [""]
  resources: ["pods"]
  verbs: ["get", "list", "watch"]
- apiGroups: [""]
  resources: ["nodes"]
  verbs: ["get"]

Config params

split_event_size int default=1000000

Docker splits long logs by 16kb chunks. The plugin joins them back, but if an event is longer than this value in bytes, it will be split after all.

Due to the optimization process it's not a strict rule. Events may be split even if they won't exceed the limit.

deleted_pods_cache_size int default=10000

How many entries for deleted pods should be stored in the cache

allowed_pod_labels []string

If set, it defines which pod labels to add to the event, others will be ignored.

allowed_node_labels []string

If set, it defines which node labels to add to the event, others will be ignored.

only_node bool default=false

Skips retrieving Kubernetes meta information using Kubernetes API and adds only k8s_node field.

watching_dir string default=/var/log/containers

Kubernetes dir with container logs. It's like watching_dir parameter from file plugin config.

offsets_file string required

The filename to store offsets of processed files. It's like offsets_file parameter from file plugin config.

file_config file.Config

Under the hood this plugin uses file plugin to collect logs from files. So you can change any file plugin config parameter using file_config section. Check out an example.

meta cfg.MetaTemplates

K8sMeta params

Add meta information to an event (look at Meta params) Use go-template syntax

Built-in meta params

k8s_pod: {{ .pod_name }}

k8s_namespace: {{ .namespace_name }}

k8s_container: {{ .container }}

k8s_container_id: {{ .container_id }}

Example: component: '{{ index .pod.Labels "component" | default .k8s_container }}'

meta_file string

The filename to store current k8s pod metadata.

Metadata contains:

• PodMeta map[Namespace]map[PodName]map[ContainerID]*podMeta, where podMeta is a wrapper for corev1.Pod
• CriType string
• NodeLabels map[string]string

Metadata is saved once per gatherer MaintenanceInterval and is loaded only on initialization.

If parameter is empty, the metadata won't be cached.

This feature is used to reuse metadata that we have already received, in particular if the kube-apiserver is unavailable.

The plugin considers kube-apiserver to be unavailable from the moment the SetWatchErrorHandler handler is called until one of the ResourceEventHandlerFuncs handlers is called.

Meta params

pod_name - string

namespace - string

container_name - string

container_id - string

pod - k8s.io/api/core/v1.Pod
Generated using insane-doc