fix: argo-repo server on AWS start cmd, upgrade flow tweaks, docs

admin/upgrade_cluster.sh

@@ -27,9 +27,6 @@ control_plane_phase1() {
 
   # upgrade modules
   admin_job "apply_kubezero, apply_network"
-
-  echo "Checking that all pods in kube-system are running ..."
-  waitSystemPodsRunning
 }
 
 control_plane_phase2() {
@@ -55,20 +52,32 @@ backup() {
   done
 }
 
+
+control_plane_healthy() {
+  echo "Checking control-plane is healthy..."
+  kubectl rollout status ds cilium -n kube-system
+  waitSystemPodsRunning
+}
+
 #####
 
-echo "Checking that all pods in kube-system are running ..."
-waitSystemPodsRunning
+
+control_plane_healthy
 
 [ "$ARGOCD" == "true" ] && disable_argo
 
 control_plane_phase1
 
+control_plane_healthy
+
+echo "Replace controller nodes first to not stall out V1.31 controllers ... <return> to continue"
+read -r
+
 cluster_modules
 
 backup
 
-echo "Once ALL nodes, incl. workers, ALL, are running on $KUBE_VERSION, <return> to continue"
+echo "Once ALL nodes, incl. workers, ALL, are running on $KUBE_VERSION <return> to continue"
 read -r
 
 control_plane_phase2

charts/kubezero/templates/argo.yaml

@@ -45,7 +45,7 @@ argo-cd:
       - name: cmp-kubezero-git-sync
         image: '{{ "{{" }} default .Values.global.image.repository .Values.repoServer.image.repository {{ "}}" }}:{{ "{{" }} default (include "argo-cd.defaultTag" .) .Values.repoServer.image.tag {{ "}}" }}'
         imagePullPolicy: '{{ "{{" }} default .Values.global.image.imagePullPolicy .Values.repoServer.image.imagePullPolicy {{ "}}" }}'
-        command: ["/var/run/argocd/argocd-cmp-server"]
+        command: ["/usr/local/bin/argocd-cmp-server"]
         env:
           {{- include "aws-iam-env" (merge (dict "roleName" "argocd-repo-server") .) | nindent 10 }}
         volumeMounts:

docs/v1.32.md

@@ -6,12 +6,13 @@
 - ArgoCD upgraded to V3 incl. various performance optimizations, eg. ServerSideApply etc.
 
 ## Features and fixes
-- enabled PSI metrics to get better insight on root cause of performance issues
+- considerable performance improvements under load for the control-plane nodes
+- enabled PSI cgroups metrics to get better insight on root cause of performance issues
 
 ## Version upgrades
 - cilium 1.17.4
 - istio 1.26.1
-- ArgoCD 3.0.5
+- ArgoCD 3.0.9
 - Prometheus 3.4.1 / Grafana 12.0.1
 - Nvidia container toolkit 1.17.8, drivers 570.148.08, Cuda 12.8
 - AWS Neuron 2.21.37