Skip to content

Development and Maintenance Guide for the Monitoring packageπŸ“œ

Code Changes for UpdatesπŸ“œ

Monitoring is a modified/customized version of an upstream chart. The below details the steps required to update to a new version of the Monitoring package.

  1. Navigate to the upstream chart repo and folder and find the tag that corresponds with the new chart version for this image update. For example, if updating the Prometheus image to 2.31.1 you would check the chart values and switch Gitlab tags (naming syntax is kube-prometheus-stack-*) until you find the latest chart version that uses the 2.31.1 Prometheus image. Start with the newest chart version to make sure we get the latest patches.

  2. Validate that the other images required by this chart version are ready and available in Ironbank. Look at the previous releases images.txt to get the list of images and search through the values file to find what the new versions are.

  3. Checkout the renovate/ironbank branch. This branch will already have most of the updates you need for the images.

  4. From the root of the repo run kpt pkg update chart@kube-prometheus-stack-23.1.6 --strategy alpha-git-patch replacing kube-prometheus-stack-23.1.6 with the version tag you got in step 1. You may be prompted to resolve some conflicts - choose what makes sense (if there are BB additions/changes keep them, if there are upstream additions/changes keep them).

  5. See the Big Bang Modifications section below for the specific changes that need to be made to the chart/Chart.yaml and chart/values.yaml files.

  6. Modify the version in Chart.yaml - you will want to append -bb.0 to the chart version from upstream.

  7. Update CHANGELOG.md adding an entry for the new version and noting all changes (at minimum should include Updated Monitoring chart to x.x.x and Updated image versions to latest in IB (P: x.x.x G: x.x.x A: x.x.x) with the versions for Prometheus, Grafana, Alertmanager).

  8. Update dependencies and binaries using helm dependency update ./chart

    • Pull assets and commit the binaries as well as the Chart.lock file that was generated.
    • If the prometheus/snmp_exporter image is being updated in Chart.yaml:
    • Check the upstream prometheus-snmp-exporter Chart.yaml file to see if there is a new chart released with the new image update

      • If a new chart exists with the new image, from the root of the repo run the following command:

        1. Run a KPT package update

          kpt pkg update chart/deps/prometheus-snmp-exporter@prometheus-snmp-exporter-<version> --strategy alpha-git-patch
          
        2. Update the file://./deps/prometheus-snmp-exporter chart version in chart/Chart.yaml and the image version in chart/values.yaml.

        3. Consult the Big Bang Modifications section for the specific changes that need to be made to the chart/deps/prometheus-snmp-exporter/values.yaml and chart/deps/prometheus-snmp-exporter/templates/deployment.yaml files.

        4. Last, update dependencies and binaries using helm dependency update ./chart.

        Note: Any time any file in the chart/deps/prometheus-snmp-exporter directory (or a sub-directory thereof) is changed, you must run helm dependency update ./chart to rebuild chart/charts/prometheus-snmp-exporter-<version>.tgz. Failure to do so will result in your changes not being included in deployments.

      • Otherwise (if a new chart does not exist with the new image), skip this image update (i.e. revert it from Chart.yaml because Renovate is trying to jump ahead) and continue to Step 9.

  9. Generate the README.md updates by following the guide in gluon.

  10. Push up your changes, validate that CI passes. If there are any failures follow the information in the pipeline to make the necessary updates and reach out to the team if needed.

  11. As part of your MR that modifies bigbang packages, you should modify the bigbang bigbang/tests/test-values.yaml against your branch for the CI/CD MR testing by enabling your packages.

    • To do this, at a minimum, you will need to follow the instructions at bigbang/docs/developer/test-package-against-bb.md with changes for Monitoring enabled (the below is a reference, actual changes could be more depending on what changes where made to Monitoring in the pakcage MR).

test-values.yamlπŸ“œ

```yaml
monitoring:
  enabled: true
  git:
    tag: null
    branch: <my-package-branch-that-needs-testing>
  values:
    istio:
      hardened:
        enabled: true
  ### Additional compononents of Monitoring should be changed to reflect testing changes introduced in the package MR
```
  1. Perform the steps below for manual testing. CI provides a good set of basic smoke tests but it is beneficial to run some additional checks.

Manual Testing for UpdatesπŸ“œ

NOTE: For these testing steps it is good to do them on both a clean install and an upgrade. For clean install, point monitoring to your branch. For an upgrade do an install with monitoring pointing to the latest tag, then perform a helm upgrade with monitoring pointing to your branch.

The following overrides can be used for a bare minimum monitoring deployment:

istio:
  enabled: true
istioOperator:
  enabled: true
monitoring:
  enabled: true
  git:
    tag: null
    branch: "renovate/ironbank"
  sso:
    enabled: true
    prometheus:
      client_id: platform1_a8604cc9-f5e9-4656-802d-d05624370245_bb8-prometheus
    alertmanager:
      client_id: platform1_a8604cc9-f5e9-4656-802d-d05624370245_bb8-alertmanager
grafana:
  sso:
    enabled: true
    grafana:
      client_id: platform1_a8604cc9-f5e9-4656-802d-d05624370245_bb8-grafana
      scopes: "openid Grafana"
jaeger:
  enabled: true
  sso:
    enabled: true
    client_id: platform1_a8604cc9-f5e9-4656-802d-d05624370245_bb8-jaeger
kiali:
  enabled: true
  sso:
    enabled: true
    client_id: platform1_a8604cc9-f5e9-4656-802d-d05624370245_bb8-kiali

addons:
  authservice:
    enabled: true

Testing Steps:

  • Login with SSO to Prometheus (if you are not prompted for an SSO login, this could indicate a problem with the authservice connection), check the Status -> Targets page and validate that all targets show as up
  • Login with SSO to Alertmanager and validate that alerts are firing (if the main page shows no alert groups check the Prometheus logs and see if there are errors with that connection)
  • Login with SSO to Grafana and take a look at some dashboards, validate that data is loaded
  • Login to Kiali and go to applications, pick monitoring namespace and prometheus for the application, validate that there is data in some of the inbound/outbound metrics fields - also validate Kiali is showing no red bells on the top bar (this could indicate connection issues with Prometheus/Grafana)

When in doubt with any testing or upgrade steps ask one of the CODEOWNERS for assistance.

Upstream changes needed for Big BangπŸ“œ

Due to how Big Bang is making use of Monitoring, there were values and chart changes that needed to be made.

This provides a log of these changes to make updates from upstream faster.

Big Bang ModificationsπŸ“œ

chart/Chart.yamlπŸ“œ

  • Line 75-77: Ensure that condition: prometheus.enabled is set for prometheus as in the following example:
- name: prometheus
  image: registry1.dso.mil/ironbank/opensource/prometheus/prometheus:vX.Y.Z
  condition: prometheus.enabled

chart/values.yamlπŸ“œ

  • Line 138: Ensure nameOverride is set to chart/values.yaml to keep resource names from changing
nameOverride: "kube-prometheus-stack"
  • Line 418: Ensure that alertmanager.serviceAccount.automountServiceAccountToken: false is set.

  • Line 901: Ensure that alertmanager.alertmanagerSpec.externalUrl is set.

## The external URL the Alertmanager instances will be available under. This is necessary to generate correct URLs. This is necessary if Alertmanager is not served from root of a DNS name. string  false
##
externalUrl: "https://alertmanager.{{ .Values.domain }}"
  • Line 1454-1487: We want to ensure that grafana.persistence.enabled=false and initChownData is using a registry1 ubiX-minimal image:
grafana:
  ...
  persistence:
    type: pvc
    enabled: false
    # storageClassName: default
    accessModes:
      - ReadWriteOnce
    size: 10Gi
    # annotations: {}

  initChownData:
    ## If false, data ownership will not be reset at startup
    ## This allows the prometheus-server to be run with an arbitrary user
    ##
    enabled: false

    ## initChownData container image
    ##
    image:
      repository: registry1.dso.mil/ironbank/redhat/ubi/9-minimal
      tag: "9.4"
      sha: ""
      pullPolicy: IfNotPresent

    ## initChownData resource requests and limits
    ## Ref: http://kubernetes.io/docs/user-guide/compute-resources/
    ##
    resources:
      limits:
        cpu: 100m
        memory: 128Mi
      requests:
        cpu: 100m
        memory: 128Mi
  • Line 2505: Ensure that prometheus-node-exporter.hostPID is set to false to resolve OPA violations with the prometheus node exporter daemonset:
prometheus-node-exporter:
  ...
  hostPID: false
  • Line 2585: Ensure that the snmpExporter configuration is present and that the snmpExporter.image.tag and snmpExporter.configmapReload.image.tag are set to the intended versions. Consult the upstream prometheus-snmp-exporter chart version for the correct versions. The following is an example of the configuration block for the SNMP exporter:
## Deploy SNMP exporter as a deployment to all nodes
##
snmpExporter:
  enabled: false

## Configuration for prometheus-snmp-exporter sub-chart
##
  image:
    repository: registry1.dso.mil/ironbank/opensource/prometheus/snmp_exporter
    tag: v0.26.0

  imagePullSecrets:
    - name: private-registry

  configmapReload:
    image:
      repository: registry1.dso.mil/ironbank/opensource/prometheus-operator/prometheus-config-reloader
      tag: v0.74.0

  ## Security context to be added to snmp-exporter pods
  securityContext:
    runAsNonRoot: true
    runAsUser: 1001
    runAsGroup: 1001
    fsGroup: 1001

  ## Security context to be added to snmp-exporter containers
  containerSecurityContext:
    runAsGroup: 1001
    runAsNonRoot: true
    runAsUser: 1001
    capabilities:
      drop:
        - ALL

  # Enable this if you're using https://github.com/prometheus-operator/prometheus-operator
  # A service monitor will be created per each item in serviceMonitor.params[]
  serviceMonitor:
    enabled: true
  • Line 3025: Ensure prometheusOperator.clusterDomain: "cluster.local" is set.

  • Line 3187-3193: Ensure that prometheusOperator.resources is set to the following:

resources:
  limits:
    cpu: 200m
    memory: 512Mi
  requests:
    cpu: 200m
    memory: 512Mi
  • Line 3298/3324: Ensure that the prometheusOperator.image.tag and prometheusOperator.prometheusConfigReloader.image.tag values are not ahead of the actual appVersion in Chart.yaml. You need to check values.yaml and Chart.yaml for unintended changes. The bot will try to jump ahead.

  • Line 3997: Ensure that prometheus.prometheusSpec.externalUrl is set.

## External URL at which Prometheus will be reachable.
##
externalUrl: "https://prometheus.{{ .Values.domain }}"

chart/deps/prometheus-snmp-exporter/values.yamlπŸ“œ

  • Line 1: Ensure that nameOverride is set to prometheus-snmp-exporter to keep resource names from changing due to the use of the snmpExporter alias.
nameOverride: "prometheus-snmp-exporter"

chart/deps/prometheus-snmp-exporter/templates/deployment.yamlπŸ“œ

  • Line 96-99: To comply with Kyverno policies, we alter the upstream chart to pass containerSecurityContext into the configmap-reload container. This is done by adding the following block to the configmap-reload container:
{{- if .Values.containerSecurityContext }}
          securityContext:
{{ toYaml .Values.containerSecurityContext | indent 12 }}
        {{- end }}

Last update: 2024-07-15 by Wesley Griffing