fix: update our galera cluster manifest to run the latest stable

This change is updating the mariadb-galera manifest to use the latest images
and system config. This is being done to help resolve some of the deadlock
issues being experienced when the cluster is under load.

This change also brings in a LOT of fixes into the operator, which should
help make our platform run better, generally.

Signed-off-by: Kevin Carter <[email protected]>

kustomize/mariadb-cluster/base/mariadb-galera.yaml

@@ -1,4 +1,4 @@
-apiVersion: mariadb.mmontes.io/v1alpha1
+apiVersion: k8s.mariadb.com/v1alpha1
 kind: MariaDB
 metadata:
   name: mariadb-galera
@@ -7,47 +7,73 @@ spec:
   rootPasswordSecretKeyRef:
     name: mariadb
     key: root-password
-
-  database: mariadb
   username: mariadb
-  passwordSecretKeyRef:
-    name: mariadb
-    key: password
-
-  image: mariadb:11.0.3
+  database: mariadb
 
-  port: 3306
+  storage:
+    size: 1Gi
+    storageClassName: general
+    resizeInUseVolumes: true
+    waitForVolumeResize: true
+    volumeClaimTemplate:
+      accessModes:
+      - ReadWriteOnce
+      resources:
+        requests:
+          storage: 1Gi
+      storageClassName: general
 
   replicas: 3
-  podSecurityContext:
-    runAsUser: 0
+
+  # point to an existing MaxScale instance. Doing this will delegate tasks such as primary failover to MaxScale.
+  # maxScaleRef:
+  #   name: maxscale
+
+  # provision a MaxScale instance and set 'spec.maxScaleRef' automatically.
+  maxScale:
+    enabled: false
+
+    kubernetesService:
+      type: LoadBalancer
+      annotations:
+        metallb.universe.tf/address-pool: primary
+
+    connection:
+      secretName: mxs-galera-conn
+      port: 3306
+
   galera:
     enabled: true
     primary:
       podIndex: 0
       automaticFailover: true
     sst: mariabackup
-    replicaThreads: 4
+    availableWhenDonor: false
+    galeraLibPath: /usr/lib/galera/libgalera_smm.so
+    replicaThreads: 1
     agent:
-      image: ghcr.io/mariadb-operator/agent:v0.0.3
+      image: ghcr.io/mariadb-operator/mariadb-operator:v0.0.26
       port: 5555
       kubernetesAuth:
         enabled: true
-      gracefulShutdownTimeout: 5s
+      gracefulShutdownTimeout: 1s
     recovery:
       enabled: true
-      clusterHealthyTimeout: 3m
+      minClusterSize: 50%
+      clusterHealthyTimeout: 30s
       clusterBootstrapTimeout: 10m
-      podRecoveryTimeout: 5m
-      podSyncTimeout: 5m
+      podRecoveryTimeout: 3m
+      podSyncTimeout: 3m
     initContainer:
-      image: ghcr.io/mariadb-operator/init:v0.0.6
-    volumeClaimTemplate:
-      resources:
-        requests:
-          storage: 10Gi
-      accessModes:
-        - ReadWriteOnce
+      image: ghcr.io/mariadb-operator/mariadb-operator:v0.0.26
+    config:
+      reuseStorageVolume: false
+      volumeClaimTemplate:
+        resources:
+          requests:
+            storage: 300Mi
+        accessModes:
+          - ReadWriteOnce
 
   service:
     type: LoadBalancer
@@ -77,17 +103,15 @@ spec:
       key: dsn
 
   affinity:
-    podAntiAffinity:
-      requiredDuringSchedulingIgnoredDuringExecution:
-        - topologyKey: "kubernetes.io/hostname"
+    enableAntiAffinity: true
 
   tolerations:
-    - key: "mariadb.mmontes.io/ha"
+    - key: "k8s.mariadb.com/ha"
       operator: "Exists"
       effect: "NoSchedule"
 
   podDisruptionBudget:
-    maxUnavailable: 66%
+    maxUnavailable: 33%
 
   updateStrategy:
     type: RollingUpdate
@@ -105,13 +129,9 @@ spec:
 
   resources:
     requests:
-      memory: 256Mi
+      memory: 128Mi
     limits:
-      memory: 16Gi
+      memory: 1Gi
 
-  volumeClaimTemplate:
-    resources:
-      requests:
-        storage: 10Gi
-    accessModes:
-      - ReadWriteOnce
+  metrics:
+    enabled: true