feat: prefetching functionality via cronjob to keep client cache warm (#145)

Co-authored-by: 1602077 <jack.charlie.munday@cern.ch>

Author: 1602077

deployments/docker/Dockerfile

@@ -15,6 +15,7 @@ RUN if [[ "${TARGETARCH}" == "amd64" ]]; then \
     https://ecsft.cern.ch/dist/cvmfs/cvmfs-${REL_VERSION}/cvmfs-libs-${PKG_VERSION}.el9.${FRIENDLY_ARCH_NAME}.rpm \
     https://ecsft.cern.ch/dist/cvmfs/cvmfs-${REL_VERSION}/cvmfs-fuse3-${PKG_VERSION}.el9.${FRIENDLY_ARCH_NAME}.rpm \
     https://ecsft.cern.ch/dist/cvmfs/cvmfs-config/cvmfs-config-default-latest.noarch.rpm \
+    cronie \
     && dnf --enablerepo=* clean all \
     && rm -rf /var/cache/yum /var/cache/dnf
 

deployments/helm/README.md

@@ -78,6 +78,11 @@ Alternatively, a YAML file that specifies the values of the parameters can be pr
 | `nodeplugin.nodeSelector` | Pod node selector of the nodeplugin DaemonSet.                                                                                            |
 | `nodeplugin.tolerations` | Pod tolerations of the nodeplugin DaemonSet.                                                                                               |
 | `nodeplugin.affinity` | Pod node affinity of the nodeplugin DaemonSet.                                                                                                |
+| `nodeplugin.prefetcher.enabled` | Whether to enable CVMFS-CSI prefetching jobs.                                                                                       |
+| `nodeplugin.prefetcher.plugin.image.repository` | Default container image repository for CVMFS CSI prefetching jobs.                                                  |
+| `nodeplugin.prefetcher.plugin.image.tag` | Default container image tag for CVMFS CSI prefetching jobs.                                                                |
+| `nodeplugin.prefetcher.plugin.image.pullPolicy` | Pull policy for CVMFS CSI prefetching jobs.                                                                         |
+| `nodeplugin.prefetcher.jobs` | Defintion of CVMFS-CSI prefetching tasks.                                                                                              |
 | `controllerplugin.name` | Component name for controller plugin component. Used as `component` label value and to generate Deployment name.                            |
 | `controllerplugin.podSecurityContext` | Pod-level security context for controllerplugin deployment.                                                                   |
 | `controllerplugin.plugin.image.repository` | Container image repository for CVMFS CSI controller plugin.                                                              |

deployments/helm/cvmfs-csi/templates/_helpers.tpl

@@ -12,8 +12,8 @@ We truncate at 63 chars because some Kubernetes name fields are limited to this
 If release name contains chart name it will be used as a full name.
 */}}
 {{- define "cvmfs-csi.fullname" -}}
-{{- if .Values.fullnameOverride -}}
-{{- .Values.fullnameOverride | trunc 63 | trimSuffix "-" -}}
+{{- if .Values.fullNameOverride -}}
+{{- .Values.fullNameOverride | trunc 63 | trimSuffix "-" -}}
 {{- else -}}
 {{- $name := default .Chart.Name .Values.nameOverride -}}
 {{- if contains $name .Release.Name -}}

deployments/helm/cvmfs-csi/templates/nodeplugin-daemonset.yaml

@@ -1,3 +1,5 @@
+{{- $defaultPrefetchImage := printf "%s:%s" $.Values.nodeplugin.prefetcher.image.repository ($.Values.nodeplugin.prefetcher.image.tag | default $.Chart.AppVersion) -}}
+---
 kind: DaemonSet
 apiVersion: apps/v1
 metadata:
@@ -188,6 +190,28 @@ spec:
           {{- with .Values.nodeplugin.singlemount.resources }}
           resources: {{ toYaml . | nindent 12 }}
           {{- end }}
+        {{- if .Values.nodeplugin.prefetcher.enabled }}
+        {{- range .Values.nodeplugin.prefetcher.jobs }}
+        - name: prefetcher-{{ .name }}
+          image: {{ .image | default $defaultPrefetchImage }}
+          # Launch crond with all debug flags (-x).
+          command: ["/usr/sbin/crond", "-n", "-s", "-m", "off", "-x", "load,proc,sch"]
+          securityContext:
+            privileged: true
+          volumeMounts:
+            - name: autofs-root
+              mountPath: /cvmfs
+              mountPropagation: Bidirectional
+            - name: prefetcher-{{ .name }}
+              mountPath: "/etc/crontab"
+              subPath: "cron.d"
+              readOnly: true
+            - name: prefetcher-{{ .name }}
+              mountPath: "/etc/cvmfs-csi/prefetcher/entrypoint.sh"
+              subPath: "entrypoint.sh"
+              readOnly: true
+        {{- end }}
+        {{- end }}
       volumes:
         - name: socket-dir
           hostPath:
@@ -225,6 +249,13 @@ spec:
         - name: cvmfs-aliencache
           {{- toYaml .Values.cache.alien.volumeSpec | nindent 10 }}
         {{- end }}
+        {{- if .Values.nodeplugin.prefetcher.enabled }}
+        {{- range .Values.nodeplugin.prefetcher.jobs }}
+        - name: prefetcher-{{ .name }}
+          configMap:
+            name: {{ include "cvmfs-csi.fullname" $ }}-prefetcher-{{ .name }}
+        {{- end }}
+        {{- end }}
         {{- with .Values.nodeplugin.extraVolumes }}
         {{- toYaml . | nindent 8 }}
         {{- end }}

deployments/helm/cvmfs-csi/templates/prefetcher-configmaps.yaml

@@ -0,0 +1,23 @@
+{{ if .Values.nodeplugin.prefetcher.enabled }}
+{{- range .Values.nodeplugin.prefetcher.jobs }}
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: {{ include "cvmfs-csi.fullname" $ }}-prefetcher-{{ .name }}
+  labels: 
+    {{- include "cvmfs-csi.common.labels" $ | nindent 4 }}
+    job: {{ .name }}
+data:
+  # Entrypoint for cvmfs-csi prefetcher job mounted at 
+  # /etc/cvmfs-csi/prefetcher/entrypoint.sh.
+  entrypoint.sh: {{ toYaml .script | nindent 4 }}
+
+  # Crontab configuration for entrypoint.sh.
+  cron.d: |
+    SHELL=/bin/bash
+    PATH=/sbin:/bin:/usr/sbin:/usr/bin
+    MAILTO=""
+    {{ .schedule }} "bash /etc/cvmfs-csi/prefetcher/entrypoint.sh"
+{{- end }}
+{{- end }}

deployments/helm/cvmfs-csi/values.yaml

@@ -105,6 +105,25 @@ nodeplugin:
       - name: etc-cvmfs-config-d
         mountPath: /etc/cvmfs/config.d
 
+  # Prefetch cvmfs repos on a schedule to keep client cache warm.
+  prefetcher:
+    enabled: false
+    # Image defines the default image for prefetching jobs.
+    # This can be optionally overridden per prefetch job.
+    image:
+      repository: registry.cern.ch/kubernetes/cvmfs-csi
+      tag: "" # If no tag specified default to Chart AppVersion.
+      pullPolicy: IfNotPresent
+    # Jobs defines the pre-fetching tasks for keeping the client cache warm.
+    jobs: {}
+      # - name: example-job # Names should be unique and confirm to k8s resource naming specification.
+      #   # image: "" # Optional if not specified defaults to prefetcher.image. Image must include crond.
+      #   schedule: "* * * * *"
+      #   script: |-
+      #     #!/bin/bash
+      #     echo -e "prefetching atlas.cern.ch"
+      #     ls /cvmfs/atlas.cern.ch
+
   # automount-runner image and container resources specs.
   singlemount:
     image: