fix: Release 1.22.8, minor upgrade fixes, bump prometheus max. memory

Dockerfile

@@ -16,6 +16,7 @@ RUN cd /etc/apk/keys && \
       cri-tools@kubezero \
       kubeadm@kubezero~=${KUBE_VERSION} \
       kubectl@kubezero~=${KUBE_VERSION} \
+      etcdhelper@kubezero \
       etcd-ctl@testing \
       restic@testing \
       helm@testing

charts/kubezero-metrics/README.md

@@ -160,7 +160,7 @@ Kubernetes: `>= 1.20.0`
 | kube-prometheus-stack.prometheus.prometheusSpec.logFormat | string | `"json"` |  |
 | kube-prometheus-stack.prometheus.prometheusSpec.podMonitorSelectorNilUsesHelmValues | bool | `false` |  |
 | kube-prometheus-stack.prometheus.prometheusSpec.portName | string | `"http-prometheus"` |  |
-| kube-prometheus-stack.prometheus.prometheusSpec.resources.limits.memory | string | `"3Gi"` |  |
+| kube-prometheus-stack.prometheus.prometheusSpec.resources.limits.memory | string | `"4Gi"` |  |
 | kube-prometheus-stack.prometheus.prometheusSpec.resources.requests.cpu | string | `"500m"` |  |
 | kube-prometheus-stack.prometheus.prometheusSpec.resources.requests.memory | string | `"512Mi"` |  |
 | kube-prometheus-stack.prometheus.prometheusSpec.retention | string | `"8d"` |  |

charts/kubezero-metrics/values.yaml

@@ -108,7 +108,7 @@ kube-prometheus-stack:
           memory: 512Mi
           cpu: 500m
         limits:
-          memory: 3Gi
+          memory: 4Gi
           # cpu: "1000m"
 
       walCompression: true

charts/kubezero/README.md

@@ -26,14 +26,14 @@ Kubernetes: `>= 1.20.0`
 |-----|------|---------|-------------|
 | HighAvailableControlplane | bool | `false` |  |
 | addons.enabled | bool | `false` |  |
-| addons.targetRevision | string | `"0.5.2"` |  |
+| addons.targetRevision | string | `"0.5.3"` |  |
 | argocd.enabled | bool | `false` |  |
 | argocd.istio.enabled | bool | `false` |  |
 | argocd.namespace | string | `"argocd"` |  |
 | argocd.targetRevision | string | `"0.10.1"` |  |
 | cert-manager.enabled | bool | `false` |  |
 | cert-manager.namespace | string | `"cert-manager"` |  |
-| cert-manager.targetRevision | string | `"0.9.0"` |  |
+| cert-manager.targetRevision | string | `"0.9.1"` |  |
 | istio-ingress.chart | string | `"kubezero-istio-gateway"` |  |
 | istio-ingress.enabled | bool | `false` |  |
 | istio-ingress.gateway.service | object | `{}` |  |

charts/kubezero/values.yaml

@@ -8,7 +8,7 @@ HighAvailableControlplane: false
 
 addons:
   enabled: false
-  targetRevision: 0.5.2
+  targetRevision: 0.5.3
 
 network:
   enabled: false
@@ -18,7 +18,7 @@ network:
 cert-manager:
   enabled: false
   namespace: cert-manager
-  targetRevision: 0.9.0
+  targetRevision: 0.9.1
 
 storage:
   enabled: false

docs/inf1-test.yaml

@@ -0,0 +1,27 @@
+apiVersion: v1
+kind: Pod
+metadata:
+  labels:
+    app: neuron-top
+  name: neuron-top
+  namespace: default
+spec:
+  containers:
+  - image: public.ecr.aws/zero-downtime/dumpster:latest
+    command:
+    - "bash"
+    - "-c"
+    - "/opt/aws/neuron/bin/neuron-ls && sleep 3600"
+    imagePullPolicy: IfNotPresent
+    name: neuron-top
+    resources:
+      limits:
+        #hugepages-2Mi: 256Mi
+        aws.amazon.com/neuron: 1
+      requests:
+        memory: 1024Mi
+  tolerations:
+  - effect: NoSchedule
+    key: kubezero-workergroup
+    operator: Equal
+    value: public

releases/v1.22/kubezero.sh

@@ -186,7 +186,7 @@ elif [[ "$1" =~ "^(bootstrap|restore|join)$" ]]; then
 
   render_kubeadm
 
-  # Esnure clean slate if bootstrap, restore PKI otherwise
+  # Ensure clean slate if bootstrap, restore PKI otherwise
   if [[ "$1" =~ "^(bootstrap)$" ]]; then
     rm -rf ${HOSTFS}/var/lib/etcd/member
 
@@ -235,6 +235,11 @@ elif [[ "$1" =~ "^(bootstrap|restore|join)$" ]]; then
   _kubeadm init phase kubeconfig all
 
   if [[ "$1" =~ "^(join)$" ]]; then
+    # Delete any former self in case forseti did not delete yet
+    kubectl delete node ${NODENAME} --wait=true || true
+    # Wait for all pods to be deleted otherwise we end up with stale pods eg. kube-proxy and all goes to ....
+    kubectl delete pods -n kube-system --field-selector spec.nodeName=${NODENAME}
+
     # get current running etcd pods for etcdctl commands
     while true; do
       etcd_endpoints=$(kubectl get pods -n kube-system -l component=etcd -o yaml | \
@@ -273,9 +278,6 @@ elif [[ "$1" =~ "^(bootstrap|restore|join)$" ]]; then
       | .etcd.initialCluster = strenv(ETCD_INITIAL_CLUSTER)
       ' ${HOSTFS}/etc/kubernetes/kubezero.yaml
     render_kubeadm
-
-    # Delete any former self in case forseti did not delete yet
-    kubectl delete node ${NODENAME} --wait=true || true
   fi
 
   # Generate our custom etcd yaml
@@ -290,6 +292,19 @@ elif [[ "$1" =~ "^(bootstrap|restore|join)$" ]]; then
   echo "Waiting for Kubernetes API to be online ..."
   retry 0 5 30 kubectl cluster-info --request-timeout 3 >/dev/null
 
+  # Update providerID as underlying VM changed during restore
+  if [[ "$1" =~ "^(restore)$" ]]; then
+    PROVIDER_ID=$(yq eval '.providerID' ${HOSTFS}/etc/kubernetes/kubezero.yaml)
+    if [ -n "$PROVIDER_ID" ]; then
+      etcdhelper \
+        -cacert ${HOSTFS}/etc/kubernetes/pki/etcd/ca.crt \
+        -cert ${HOSTFS}/etc/kubernetes/pki/etcd/server.crt \
+        -key ${HOSTFS}/etc/kubernetes/pki/etcd/server.key \
+        -endpoint https://${ETCD_NODENAME}:2379 \
+        change-provider-id ${NODENAME} $PROVIDER_ID
+    fi
+  fi
+
   if [[ ! "$1" =~ "^(join)$" ]]; then
     _kubeadm init phase upload-config all
     _kubeadm init phase upload-certs --skip-certificate-key-print

releases/v1.22/migrate_argo.py

@@ -15,7 +15,7 @@ parser = argparse.ArgumentParser(description="Update Route53 entries")
 parser.add_argument(
     "--version",
     dest="version",
-		default="1.22.8-7",
+		default="1.22.8-8",
     action="store",
     required=False,
     help="Update KubeZero version",