chore(cnpg): add retentionPolicy option for backups

Smana · Dec 5, 2024 · b6baf9b · b6baf9b
1 parent d935234
commit b6baf9b
Show file tree

Hide file tree

Showing 40 changed files with 209 additions and 12,661 deletions.
diff --git a/clusters/mycluster-0/flux-system/gotk-components.yaml b/clusters/mycluster-0/flux-system/gotk-components.yaml
diff --git a/clusters/mycluster-0/flux-system/gotk-sync.yaml b/clusters/mycluster-0/flux-system/gotk-sync.yaml
diff --git a/clusters/mycluster-0/flux-system/kustomization.yaml b/clusters/mycluster-0/flux-system/kustomization.yaml
diff --git a/clusters/mycluster-0/observability.yaml b/clusters/mycluster-0/observability.yaml
@@ -25,37 +25,6 @@ spec:
       name: victoria-metrics-k8s-stack
       namespace: observability
 ---
-apiVersion: kustomize.toolkit.fluxcd.io/v1beta2
-kind: Kustomization
-metadata:
-  name: observability-grafana-oncall
-  namespace: flux-system
-spec:
-  interval: 3m0s
-  timeout: 10m0s
-  path: ./observability/mycluster-0/grafana-oncall
-  prune: true
-  sourceRef:
-    kind: GitRepository
-    name: flux-system
-  postBuild:
-    substitute:
-      domain_name: "cloud.ogenki.io"
-    substituteFrom:
-      - kind: ConfigMap
-        name: eks-mycluster-0-vars
-  dependsOn:
-    - name: observability-victoria-metrics-k8s-stack
-  healthChecks:
-    - apiVersion: helm.toolkit.fluxcd.io/v2
-      kind: HelmRelease
-      name: oncall
-      namespace: observability
-    - apiVersion: helm.toolkit.fluxcd.io/v2
-      kind: HelmRelease
-      name: oncall-valkey
-      namespace: observability
----
 # The Grafana folders must be deployed first to avoid errors when dashboards are created
 apiVersion: kustomize.toolkit.fluxcd.io/v1beta2
 kind: Kustomization
@@ -124,3 +93,35 @@ spec:
   #   kind: HelmRelease
   #   name: vector-agent
   #   namespace: observability
+# Don't want to deploy OnCall every time, it's a bit heavy
+# ---
+# apiVersion: kustomize.toolkit.fluxcd.io/v1beta2
+# kind: Kustomization
+# metadata:
+#   name: observability-grafana-oncall
+#   namespace: flux-system
+# spec:
+#   interval: 3m0s
+#   timeout: 10m0s
+#   path: ./observability/base/grafana-oncall
+#   prune: true
+#   sourceRef:
+#     kind: GitRepository
+#     name: flux-system
+#   postBuild:
+#     substitute:
+#       domain_name: "cloud.ogenki.io"
+#     substituteFrom:
+#       - kind: ConfigMap
+#         name: eks-mycluster-0-vars
+#   dependsOn:
+#     - name: observability-victoria-metrics-k8s-stack
+#   healthChecks:
+#     - apiVersion: helm.toolkit.fluxcd.io/v2
+#       kind: HelmRelease
+#       name: oncall
+#       namespace: observability
+#     - apiVersion: helm.toolkit.fluxcd.io/v2
+#       kind: HelmRelease
+#       name: oncall-valkey
+#       namespace: observability
diff --git a/crds/base/kustomization-rabbitmq-cluster-operator.yaml b/crds/base/kustomization-rabbitmq-cluster-operator.yaml
diff --git a/crds/base/kustomization.yaml b/crds/base/kustomization.yaml
@@ -10,5 +10,4 @@ resources:
   - kustomization-grafana-operator.yaml
   - kustomization-kyverno.yaml
   - kustomization-cloudnative-pg.yaml
-  - kustomization-rabbitmq-cluster-operator.yaml
   - kustomization-victoria-metrics-operator.yaml
diff --git a/dagger/go.mod b/dagger/go.mod
@@ -33,14 +33,14 @@ require (
 	go.opentelemetry.io/otel/sdk/metric v1.32.0
 	go.opentelemetry.io/otel/trace v1.32.0
 	go.opentelemetry.io/proto/otlp v1.4.0
-	golang.org/x/exp v0.0.0-20241108190413-2d47ceb2692f
-	golang.org/x/net v0.31.0 // indirect
-	golang.org/x/sync v0.9.0
-	golang.org/x/sys v0.27.0 // indirect
-	golang.org/x/text v0.20.0 // indirect
+	golang.org/x/exp v0.0.0-20241204233417-43b7b7cde48d
+	golang.org/x/net v0.32.0 // indirect
+	golang.org/x/sync v0.10.0
+	golang.org/x/sys v0.28.0 // indirect
+	golang.org/x/text v0.21.0 // indirect
 	google.golang.org/genproto/googleapis/api v0.0.0-20241202173237-19429a94021a // indirect
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20241202173237-19429a94021a // indirect
-	google.golang.org/grpc v1.68.0
+	google.golang.org/grpc v1.68.1
 	google.golang.org/protobuf v1.35.2 // indirect
 )
 

diff --git a/dagger/go.sum b/dagger/go.sum
@@ -75,22 +75,22 @@ go.opentelemetry.io/proto/otlp v1.4.0 h1:TA9WRvW6zMwP+Ssb6fLoUIuirti1gGbP28GcKG1
 go.opentelemetry.io/proto/otlp v1.4.0/go.mod h1:PPBWZIP98o2ElSqI35IHfu7hIhSwvc5N38Jw8pXuGFY=
 go.uber.org/goleak v1.3.0 h1:2K3zAYmnTNqV73imy9J1T3WC+gmCePx2hEGkimedGto=
 go.uber.org/goleak v1.3.0/go.mod h1:CoHD4mav9JJNrW/WLlf7HGZPjdw8EucARQHekz1X6bE=
-golang.org/x/exp v0.0.0-20241108190413-2d47ceb2692f h1:XdNn9LlyWAhLVp6P/i8QYBW+hlyhrhei9uErw2B5GJo=
-golang.org/x/exp v0.0.0-20241108190413-2d47ceb2692f/go.mod h1:D5SMRVC3C2/4+F/DB1wZsLRnSNimn2Sp/NPsCrsv8ak=
-golang.org/x/net v0.31.0 h1:68CPQngjLL0r2AlUKiSxtQFKvzRVbnzLwMUn5SzcLHo=
-golang.org/x/net v0.31.0/go.mod h1:P4fl1q7dY2hnZFxEk4pPSkDHF+QqjitcnDjUQyMM+pM=
-golang.org/x/sync v0.9.0 h1:fEo0HyrW1GIgZdpbhCRO0PkJajUS5H9IFUztCgEo2jQ=
-golang.org/x/sync v0.9.0/go.mod h1:Czt+wKu1gCyEFDUtn0jG5QVvpJ6rzVqr5aXyt9drQfk=
-golang.org/x/sys v0.27.0 h1:wBqf8DvsY9Y/2P8gAfPDEYNuS30J4lPHJxXSb/nJZ+s=
-golang.org/x/sys v0.27.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
-golang.org/x/text v0.20.0 h1:gK/Kv2otX8gz+wn7Rmb3vT96ZwuoxnQlY+HlJVj7Qug=
-golang.org/x/text v0.20.0/go.mod h1:D4IsuqiFMhST5bX19pQ9ikHC2GsaKyk/oF+pn3ducp4=
+golang.org/x/exp v0.0.0-20241204233417-43b7b7cde48d h1:0olWaB5pg3+oychR51GUVCEsGkeCU/2JxjBgIo4f3M0=
+golang.org/x/exp v0.0.0-20241204233417-43b7b7cde48d/go.mod h1:qj5a5QZpwLU2NLQudwIN5koi3beDhSAlJwa67PuM98c=
+golang.org/x/net v0.32.0 h1:ZqPmj8Kzc+Y6e0+skZsuACbx+wzMgo5MQsJh9Qd6aYI=
+golang.org/x/net v0.32.0/go.mod h1:CwU0IoeOlnQQWJ6ioyFrfRuomB8GKF6KbYXZVyeXNfs=
+golang.org/x/sync v0.10.0 h1:3NQrjDixjgGwUOCaF8w2+VYHv0Ve/vGYSbdkTa98gmQ=
+golang.org/x/sync v0.10.0/go.mod h1:Czt+wKu1gCyEFDUtn0jG5QVvpJ6rzVqr5aXyt9drQfk=
+golang.org/x/sys v0.28.0 h1:Fksou7UEQUWlKvIdsqzJmUmCX3cZuD2+P3XyyzwMhlA=
+golang.org/x/sys v0.28.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
+golang.org/x/text v0.21.0 h1:zyQAAkrwaneQ066sspRyJaG9VNi/YJ1NfzcGB3hZ/qo=
+golang.org/x/text v0.21.0/go.mod h1:4IBbMaMmOPCJ8SecivzSH54+73PCFmPWxNTLm+vZkEQ=
 google.golang.org/genproto/googleapis/api v0.0.0-20241202173237-19429a94021a h1:OAiGFfOiA0v9MRYsSidp3ubZaBnteRUyn3xB2ZQ5G/E=
 google.golang.org/genproto/googleapis/api v0.0.0-20241202173237-19429a94021a/go.mod h1:jehYqy3+AhJU9ve55aNOaSml7wUXjF9x6z2LcCfpAhY=
 google.golang.org/genproto/googleapis/rpc v0.0.0-20241202173237-19429a94021a h1:hgh8P4EuoxpsuKMXX/To36nOFD7vixReXgn8lPGnt+o=
 google.golang.org/genproto/googleapis/rpc v0.0.0-20241202173237-19429a94021a/go.mod h1:5uTbfoYQed2U9p3KIj2/Zzm02PYhndfdmML0qC3q3FU=
-google.golang.org/grpc v1.68.0 h1:aHQeeJbo8zAkAa3pRzrVjZlbz6uSfeOXlJNQM0RAbz0=
-google.golang.org/grpc v1.68.0/go.mod h1:fmSPC5AsjSBCK54MyHRx48kpOti1/jRfOlwEWywNjWA=
+google.golang.org/grpc v1.68.1 h1:oI5oTa11+ng8r8XMMN7jAOmWfPZWbYpCFaMUTACxkM0=
+google.golang.org/grpc v1.68.1/go.mod h1:+q1XYFJjShcqn0QZHvCyeR4CXPA+llXIeUIfIe00waw=
 google.golang.org/protobuf v1.35.2 h1:8Ar7bF+apOIoThw1EdZl0p1oWvMqTHmpA2fRTyZO8io=
 google.golang.org/protobuf v1.35.2/go.mod h1:9fA7Ob0pmnwhb644+1+CVWFRbNajQ6iRojtC/QF5bRE=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=

diff --git a/flux/sources/helmrepo-bitnami.yaml b/flux/sources/helmrepo-bitnami.yaml
@@ -4,5 +4,6 @@ metadata:
   name: bitnami
   namespace: flux-system
 spec:
+  type: "oci"
   interval: 5m
-  url: https://charts.bitnami.com/bitnami
+  url: oci://registry-1.docker.io/bitnamicharts
diff --git a/infrastructure/base/cloudnative-pg/grafana-dashboard.yaml b/infrastructure/base/cloudnative-pg/grafana-dashboard.yaml
@@ -3,11 +3,12 @@ kind: GrafanaDashboard
 metadata:
   name: databases-cloudnative-pg
 spec:
+  allowCrossNamespaceImport: true
   folderRef: "databases"
   datasources:
     - inputName: "DS_PROMETHEUS"
       datasourceName: "VictoriaMetrics"
   instanceSelector:
     matchLabels:
       dashboards: "grafana"
-  url: "https://grafana.com/api/dashboards/20417/revisions/1/download"
+  url: "https://grafana.com/api/dashboards/20417/revisions/3/download"
diff --git a/infrastructure/base/crossplane/configuration/kcl/cnpginstance/kcl.mod b/infrastructure/base/crossplane/configuration/kcl/cnpginstance/kcl.mod
@@ -1,4 +1,4 @@
 [package]
 name = "cnpginstance"
 edition = "v0.10.7"
-version = "0.0.28"
+version = "0.0.29"
diff --git a/infrastructure/base/crossplane/configuration/kcl/cnpginstance/main.k b/infrastructure/base/crossplane/configuration/kcl/cnpginstance/main.k
@@ -105,7 +105,14 @@ if oxr.spec.cnpg:
                                         s3Credentials = {
                                             inheritFromIAMRole = True
                                         }
+                                        wal = {
+                                            compression = "bzip2"
+                                        }
+                                        data = {
+                                            compression = "bzip2"
+                                        }
                                     }
+                                    retentionPolicy = oxr.spec.cnpg.backup.retentionPolicy
                                 }
                             if oxr.spec.cnpg?.createSuperuser:
                                 superuserSecret = {

diff --git a/infrastructure/base/crossplane/configuration/kcl/cnpginstance/settings-example.yaml b/infrastructure/base/crossplane/configuration/kcl/cnpginstance/settings-example.yaml
@@ -22,6 +22,7 @@ kcl_options:
             backup:
               schedule: "0 0 * * *"
               bucketName: "myname-rds-instance-backups"
+              retentionPolicy: "10d"
             initSQL: |
               CREATE EXTENSION pg_stat_statements;
               CREATE EXTENSION pg_trgm;

diff --git a/infrastructure/base/crossplane/configuration/sql-instance-composition.yaml b/infrastructure/base/crossplane/configuration/sql-instance-composition.yaml
@@ -51,7 +51,7 @@ spec:
               kind: KCLRun
               spec:
                   target: Resources
-                  source: oci://ttl.sh/ogenki-cnref/cnpginstance:v0.0.28-24h
+                  source: oci://ttl.sh/ogenki-cnref/cnpginstance:v0.0.29-24h
 
         - step: ready
           functionRef:

diff --git a/infrastructure/base/crossplane/configuration/sql-instance-definition.yaml b/infrastructure/base/crossplane/configuration/sql-instance-definition.yaml
@@ -137,6 +137,10 @@ spec:
                           description: The name of the S3 bucket to store backups.
                           type: string
                           nullable: true
+                        retentionPolicy:
+                          description: The retention policy for backups.
+                          type: string
+                          default: "15d"
                     initSQL:
                       description: SQL to run on the primary instance after creation.
                       type: string

diff --git a/infrastructure/base/rabbitmq-cluster-operator/helmrelease.yaml b/infrastructure/base/rabbitmq-cluster-operator/helmrelease.yaml
diff --git a/infrastructure/base/rabbitmq-cluster-operator/kustomization.yaml b/infrastructure/base/rabbitmq-cluster-operator/kustomization.yaml
diff --git a/infrastructure/mycluster-0/kustomization.yaml b/infrastructure/mycluster-0/kustomization.yaml
@@ -7,4 +7,3 @@ resources:
   - ../base/cloudnative-pg
   - ../base/external-dns
   - ../base/gapi
-  - ../base/rabbitmq-cluster-operator
diff --git a/observability/base/grafana-oncall/helmrelease-oncall.yaml b/observability/base/grafana-oncall/helmrelease-oncall.yaml
@@ -14,7 +14,7 @@ spec:
         kind: HelmRepository
         name: grafana
         namespace: observability
-      version: "1.13.5"
+      version: "1.13.6"
   interval: 5m0s
   timeout: 15m
   install:
@@ -87,7 +87,7 @@ spec:
       enabled: false
 
     externalRedis:
-      host: oncall-valkey-master
+      host: oncall-valkey-primary
       port: 6379
       username: default
       existingSecret: "oncall-valkey"

diff --git a/observability/base/grafana-oncall/helmrelease-rabbitmq.yaml b/observability/base/grafana-oncall/helmrelease-rabbitmq.yaml
@@ -14,7 +14,7 @@ spec:
         kind: HelmRepository
         name: bitnami
         namespace: flux-system
-      version: "15.0.6"
+      version: "15.1.0"
   interval: 5m0s
   timeout: 15m
   install:
@@ -55,6 +55,7 @@ spec:
         perObject:
           enabled: true
         detailed:
+          family: ["queue_coarse_metrics", "queue_consumer_count"]
           enabled: true
 
       prometheusRule:

diff --git a/observability/base/grafana-oncall/helmrelease-valkey.yaml b/observability/base/grafana-oncall/helmrelease-valkey.yaml
@@ -13,7 +13,7 @@ spec:
         kind: HelmRepository
         name: bitnami
         namespace: flux-system
-      version: "1.0.2"
+      version: "2.1.1"
   interval: 10m0s
   install:
     remediation:
@@ -23,7 +23,7 @@ spec:
       existingSecret: "oncall-valkey"
       existingSecretPasswordKey: "password"
 
-    master:
+    primary:
       ## Valkey master resource requests and limits
       ## ref: https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/
       ## @param master.resourcesPreset Set container resources according to one common preset (allowed values: none, nano, micro, small, medium, large, xlarge, 2xlarge). This is ignored if master.resources is set (master.resources is recommended for production).

diff --git a/observability/base/grafana-operator/dashboards/kubernetes-karpenter.yaml b/observability/base/grafana-operator/dashboards/kubernetes-karpenter.yaml
@@ -2,7 +2,9 @@ apiVersion: grafana.integreatly.org/v1beta1
 kind: GrafanaDashboard
 metadata:
   name: kubernetes-karpenter
+  namespace: infrastructure
 spec:
+  allowCrossNamespaceImport: true
   folderRef: "kubernetes"
   datasources:
     - inputName: "DS_PROMETHEUS"

diff --git a/observability/base/grafana-operator/dashboards/kubernetes-logs.yaml b/observability/base/grafana-operator/dashboards/kubernetes-logs.yaml
@@ -2,7 +2,9 @@ apiVersion: grafana.integreatly.org/v1beta1
 kind: GrafanaDashboard
 metadata:
     name: kubernetes-logs
+    namespace: infrastructure
 spec:
+    allowCrossNamespaceImport: true
     folderRef: "kubernetes"
     datasources:
         - inputName: "DS_LOKI"

diff --git a/observability/base/grafana-operator/dashboards/kubernetes-node-exporter-full.yaml b/observability/base/grafana-operator/dashboards/kubernetes-node-exporter-full.yaml
@@ -2,7 +2,9 @@ apiVersion: grafana.integreatly.org/v1beta1
 kind: GrafanaDashboard
 metadata:
   name: kubernetes-node-exporter-full
+  namespace: infrastructure
 spec:
+  allowCrossNamespaceImport: true
   folderRef: "kubernetes"
   datasources:
     - inputName: "DS_PROMETHEUS"

diff --git a/observability/base/grafana-operator/dashboards/kubernetes-quick-search-logs.yaml b/observability/base/grafana-operator/dashboards/kubernetes-quick-search-logs.yaml
@@ -2,7 +2,9 @@ apiVersion: grafana.integreatly.org/v1beta1
 kind: GrafanaDashboard
 metadata:
     name: kubernetes-quick-search-logs
+    namespace: infrastructure
 spec:
+    allowCrossNamespaceImport: true
     folderRef: "kubernetes"
     datasources:
         - inputName: "DS_LOKI"