Simplest way of getting failure notification emails from kubernetes

The simplest solution I managed to find is robusta.

It still has a bunch of unnecessary features, but with the correct configuration it's possible to disable these. As a bonus, it nicely adds some extra info to the notifications (called enrichments).

It's intended to run alongside prometheus, but it will also work without it. Here's the relevant installation guide.

The resulting deployment consist of two pods. One watching the kubernetes API for events and another for sending the notifications (+ maybe one more for sending email, if you don't have a mailserver around).

Here's a helm values skeleton for minimal setup: (replace the stuff in <>'s)

# See here for the unmodified values:
# https://github.com/robusta-dev/robusta/blob/master/helm/robusta/values.yaml

clusterName: <your-cluster>

isSmallCluster: true

# Disable all phoning home.
disableCloudRouting: true
runner:
  sendAdditionalTelemetry: false
  additional_env_vars:
      # Some telemetry is apparently enabled by default.
      # https://github.com/robusta-dev/robusta/blob/master/helm/robusta/templates/NOTES.txt
      - name: ENABLE_TELEMETRY
        value: "false"

enablePrometheusStack: false

# https://docs.robusta.dev/master/configuration/sinks/mail.html
sinksConfig:
  - mail_sink:
      name: mail_sink
      # You need to have an email server somewhere.
      mailto: "mailto://<my.domain>?smtp=<my.smtp.server>&from=<my.cluster@add.ress>&to=<my.email@add.ress"

# builtin playbooks
builtinPlaybooks:
# playbooks for non-prometheus based monitoring
  - name: "CrashLoopBackOff"
    triggers:
      - on_pod_crash_loop:
          restart_reason: "CrashLoopBackOff"
    actions:
      - report_crash_loop: {}

  - name: "ImagePullBackOff"
    triggers:
      - on_image_pull_backoff: {}
    actions:
      - image_pull_backoff_reporter: {}

  # playbooks for non-prometheus based monitoring that use prometheus for enrichment
  - name: "PodOOMKill"
    triggers:
      - on_pod_oom_killed:
          rate_limit: 3600
    actions:
    - pod_oom_killer_enricher:
        attach_logs: true
        container_memory_graph: true
        node_memory_graph: true
    stop: true
  # Prometheus trigger playbooks removed.


# Robusta UI sinks have been disabled.
enablePlatformPlaybooks: true
platformPlaybooks:
  - name: "K8sWarningEventsReport"
    triggers:
    - on_kubernetes_warning_event_create:
        exclude: ["NodeSysctlChange"]
    actions:
    - event_report: {}
    - event_resource_events: {}
    # sinks:
    #   - "robusta_ui_sink"

  - name: "IngressChangeTracking"
    triggers:
      - on_ingress_all_changes: {}
    actions:
      - resource_babysitter: {}
      - customise_finding:
          title: Ingress Changes
          aggregation_key: IngressChange
    # sinks:
    #   - "robusta_ui_sink"

  - name: "EventBasedChangeTracking"
    triggers:
      - on_kubernetes_resource_operation:
          resources: ["deployment", "replicaset", "daemonset", "statefulset", "pod", "node", "job" ]
    actions:
      - resource_events_diff: {}

  - name: "K8sJobFailure"
    triggers:
    - on_job_failure: {}
    actions:
    - create_finding:
        aggregation_key: "job_failure"
        title: "Job Failed"
    - job_info_enricher: {}
    - job_events_enricher: {}
    - job_pod_enricher: {}
    # sinks:
    #   - "robusta_ui_sink"


# Read-only setup.
# Don't try to autofix anything.
lightActions:
  - related_pods
  # - prometheus_enricher
  - add_silence
  # - delete_pod
  - delete_silence
  - get_silences
  - logs_enricher
  - pod_events_enricher
  - deployment_events_enricher
  - job_events_enricher
  - job_pod_enricher
  - get_resource_yaml
  - node_cpu_enricher
  - node_disk_analyzer
  - node_running_pods_enricher
  - node_allocatable_resources_enricher
  - node_status_enricher
  - node_graph_enricher
  - oomkilled_container_graph_enricher
  - pod_oom_killer_enricher
  - oom_killer_enricher
  - volume_analysis
  - python_profiler
  - pod_ps
  - python_memory
  - debugger_stack_trace
  - python_process_inspector
  # - prometheus_alert
  # - create_pvc_snapshot
  - resource_events_enricher
  # - delete_job
  - list_resource_names
  - node_dmesg_enricher
  - status_enricher
  # No scanning either
  # - popeye_scan
  # - krr_scan

  # - handle_alertmanager_event
  # - drain
  # - cordon
  # - uncordon
  # - rollout_restart
  # - prometheus_all_available_metrics
  # - prometheus_get_series

posted 24 days ago

CC BY-SA 4.0

Iizuki‭

1087 reputation 20 21 139 28