diff --git a/nixos/modules/services/cluster/kubernetes/addon-manager.nix b/nixos/modules/services/cluster/kubernetes/addon-manager.nix
index ad7d17c9c283..17f2dde31a71 100644
--- a/nixos/modules/services/cluster/kubernetes/addon-manager.nix
+++ b/nixos/modules/services/cluster/kubernetes/addon-manager.nix
@@ -62,50 +62,19 @@ in
       '';
     };
 
-    enable = mkEnableOption "Kubernetes addon manager";
-
-    kubeconfig = top.lib.mkKubeConfigOptions "Kubernetes addon manager";
-    bootstrapAddonsKubeconfig = top.lib.mkKubeConfigOptions "Kubernetes addon manager bootstrap";
+    enable = mkEnableOption "Whether to enable Kubernetes addon manager.";
   };
 
   ###### implementation
-  config = let
-
-    addonManagerPaths = filter (a: a != null) [
-      cfg.kubeconfig.caFile
-      cfg.kubeconfig.certFile
-      cfg.kubeconfig.keyFile
-    ];
-    bootstrapAddonsPaths = filter (a: a != null) [
-      cfg.bootstrapAddonsKubeconfig.caFile
-      cfg.bootstrapAddonsKubeconfig.certFile
-      cfg.bootstrapAddonsKubeconfig.keyFile
-    ];
-
-  in mkIf cfg.enable {
+  config = mkIf cfg.enable {
     environment.etc."kubernetes/addons".source = "${addons}/";
 
-    #TODO: Get rid of kube-addon-manager in the future for the following reasons
-    # - it is basically just a shell script wrapped around kubectl
-    # - it assumes that it is clusterAdmin or can gain clusterAdmin rights through serviceAccount
-    # - it is designed to be used with k8s system components only
-    # - it would be better with a more Nix-oriented way of managing addons
     systemd.services.kube-addon-manager = {
       description = "Kubernetes addon manager";
       wantedBy = [ "kubernetes.target" ];
-      after = [ "kube-node-online.target" ];
-      before = [ "kubernetes.target" ];
-      environment = {
-        ADDON_PATH = "/etc/kubernetes/addons/";
-        KUBECONFIG = top.lib.mkKubeConfig "kube-addon-manager" cfg.kubeconfig;
-      };
-      path = with pkgs; [ gawk kubectl ];
-      preStart = ''
-        until kubectl -n kube-system get serviceaccounts/default 2>/dev/null; do
-          echo kubectl -n kube-system get serviceaccounts/default: exit status $?
-          sleep 2
-        done
-      '';
+      after = [ "kube-apiserver.service" ];
+      environment.ADDON_PATH = "/etc/kubernetes/addons/";
+      path = [ pkgs.gawk ];
       serviceConfig = {
         Slice = "kubernetes.slice";
         ExecStart = "${top.package}/bin/kube-addons";
@@ -115,52 +84,8 @@ in
         Restart = "on-failure";
         RestartSec = 10;
       };
-      unitConfig.ConditionPathExists = addonManagerPaths;
     };
 
-    systemd.paths.kube-addon-manager = {
-      wantedBy = [ "kube-addon-manager.service" ];
-      pathConfig = {
-        PathExists = addonManagerPaths;
-        PathChanged = addonManagerPaths;
-      };
-    };
-
-    services.kubernetes.addonManager.kubeconfig.server = mkDefault top.apiserverAddress;
-
-    systemd.services.kube-addon-manager-bootstrap = mkIf (top.apiserver.enable && top.addonManager.bootstrapAddons != {}) {
-      wantedBy = [ "kube-control-plane-online.target" ];
-      after = [ "kube-apiserver.service" ];
-      before = [ "kube-control-plane-online.target" ];
-      path = [ pkgs.kubectl ];
-      environment = {
-        KUBECONFIG = top.lib.mkKubeConfig "kube-addon-manager-bootstrap" cfg.bootstrapAddonsKubeconfig;
-      };
-      preStart = with pkgs; let
-        files = mapAttrsToList (n: v: writeText "${n}.json" (builtins.toJSON v))
-          cfg.bootstrapAddons;
-      in ''
-        until kubectl auth can-i '*' '*' -q 2>/dev/null; do
-          echo kubectl auth can-i '*' '*': exit status $?
-          sleep 2
-        done
-
-        kubectl apply -f ${concatStringsSep " \\\n -f " files}
-      '';
-      script = "echo Ok";
-      unitConfig.ConditionPathExists = bootstrapAddonsPaths;
-    };
-
-    systemd.paths.kube-addon-manager-bootstrap = {
-      wantedBy = [ "kube-addon-manager-bootstrap.service" ];
-      pathConfig = {
-        PathExists = bootstrapAddonsPaths;
-        PathChanged = bootstrapAddonsPaths;
-      };
-    };
-
-    services.kubernetes.addonManager.bootstrapAddonsKubeconfig.server = mkDefault top.apiserverAddress;
-
     services.kubernetes.addonManager.bootstrapAddons = mkIf isRBACEnabled
     (let
       name = system:kube-addon-manager;
diff --git a/nixos/modules/services/cluster/kubernetes/addons/dashboard.nix b/nixos/modules/services/cluster/kubernetes/addons/dashboard.nix
index 2295694ffc74..454e7d35bc01 100644
--- a/nixos/modules/services/cluster/kubernetes/addons/dashboard.nix
+++ b/nixos/modules/services/cluster/kubernetes/addons/dashboard.nix
@@ -169,23 +169,6 @@ in {
         };
       };
 
-      kubernetes-dashboard-cm = {
-        apiVersion = "v1";
-        kind = "ConfigMap";
-        metadata = {
-          labels = {
-            k8s-app = "kubernetes-dashboard";
-            # Allows editing resource and makes sure it is created first.
-            "addonmanager.kubernetes.io/mode" = "EnsureExists";
-          };
-          name = "kubernetes-dashboard-settings";
-          namespace = "kube-system";
-        };
-      };
-    };
-
-    services.kubernetes.addonManager.bootstrapAddons = mkMerge [{
-
       kubernetes-dashboard-sa = {
         apiVersion = "v1";
         kind = "ServiceAccount";
@@ -227,9 +210,20 @@ in {
         };
         type = "Opaque";
       };
-    }
-
-    (optionalAttrs cfg.rbac.enable
+      kubernetes-dashboard-cm = {
+        apiVersion = "v1";
+        kind = "ConfigMap";
+        metadata = {
+          labels = {
+            k8s-app = "kubernetes-dashboard";
+            # Allows editing resource and makes sure it is created first.
+            "addonmanager.kubernetes.io/mode" = "EnsureExists";
+          };
+          name = "kubernetes-dashboard-settings";
+          namespace = "kube-system";
+        };
+      };
+    } // (optionalAttrs cfg.rbac.enable
       (let
         subjects = [{
           kind = "ServiceAccount";
@@ -329,6 +323,6 @@ in {
             inherit subjects;
           };
         })
-    ))];
+    ));
   };
 }
diff --git a/nixos/modules/services/cluster/kubernetes/apiserver.nix b/nixos/modules/services/cluster/kubernetes/apiserver.nix
index f293dd79f42a..33796bf2e080 100644
--- a/nixos/modules/services/cluster/kubernetes/apiserver.nix
+++ b/nixos/modules/services/cluster/kubernetes/apiserver.nix
@@ -290,32 +290,11 @@ in
   ###### implementation
   config = mkMerge [
 
-    (let
-
-      apiserverPaths = filter (a: a != null) [
-        cfg.clientCaFile
-        cfg.etcd.caFile
-        cfg.etcd.certFile
-        cfg.etcd.keyFile
-        cfg.kubeletClientCaFile
-        cfg.kubeletClientCertFile
-        cfg.kubeletClientKeyFile
-        cfg.serviceAccountKeyFile
-        cfg.tlsCertFile
-        cfg.tlsKeyFile
-      ];
-      etcdPaths = filter (a: a != null) [
-        config.services.etcd.trustedCaFile
-        config.services.etcd.certFile
-        config.services.etcd.keyFile
-      ];
-
-    in mkIf cfg.enable {
+    (mkIf cfg.enable {
         systemd.services.kube-apiserver = {
           description = "Kubernetes APIServer Service";
-          wantedBy = [ "kube-control-plane-online.target" ];
-          after = [ "certmgr.service" ];
-          before = [ "kube-control-plane-online.target" ];
+          wantedBy = [ "kubernetes.target" ];
+          after = [ "network.target" ];
           serviceConfig = {
             Slice = "kubernetes.slice";
             ExecStart = ''${top.package}/bin/kube-apiserver \
@@ -386,15 +365,6 @@ in
             Restart = "on-failure";
             RestartSec = 5;
           };
-          unitConfig.ConditionPathExists = apiserverPaths;
-        };
-
-        systemd.paths.kube-apiserver = mkIf top.apiserver.enable {
-          wantedBy = [ "kube-apiserver.service" ];
-          pathConfig = {
-            PathExists = apiserverPaths;
-            PathChanged = apiserverPaths;
-          };
         };
 
         services.etcd = {
@@ -408,18 +378,6 @@ in
           initialAdvertisePeerUrls = mkDefault ["https://${top.masterAddress}:2380"];
         };
 
-        systemd.services.etcd = {
-          unitConfig.ConditionPathExists = etcdPaths;
-        };
-
-        systemd.paths.etcd = {
-          wantedBy = [ "etcd.service" ];
-          pathConfig = {
-            PathExists = etcdPaths;
-            PathChanged = etcdPaths;
-          };
-        };
-
         services.kubernetes.addonManager.bootstrapAddons = mkIf isRBACEnabled {
 
           apiserver-kubelet-api-admin-crb = {
diff --git a/nixos/modules/services/cluster/kubernetes/controller-manager.nix b/nixos/modules/services/cluster/kubernetes/controller-manager.nix
index b94e8bd86d4c..0b73d090f241 100644
--- a/nixos/modules/services/cluster/kubernetes/controller-manager.nix
+++ b/nixos/modules/services/cluster/kubernetes/controller-manager.nix
@@ -104,31 +104,11 @@ in
   };
 
   ###### implementation
-  config = let
-
-    controllerManagerPaths = filter (a: a != null) [
-      cfg.kubeconfig.caFile
-      cfg.kubeconfig.certFile
-      cfg.kubeconfig.keyFile
-      cfg.rootCaFile
-      cfg.serviceAccountKeyFile
-      cfg.tlsCertFile
-      cfg.tlsKeyFile
-    ];
-
-  in mkIf cfg.enable {
-    systemd.services.kube-controller-manager = rec {
+  config = mkIf cfg.enable {
+    systemd.services.kube-controller-manager = {
       description = "Kubernetes Controller Manager Service";
-      wantedBy = [ "kube-control-plane-online.target" ];
+      wantedBy = [ "kubernetes.target" ];
       after = [ "kube-apiserver.service" ];
-      before = [ "kube-control-plane-online.target" ];
-      environment.KUBECONFIG = top.lib.mkKubeConfig "kube-controller-manager" cfg.kubeconfig;
-      preStart = ''
-        until kubectl auth can-i get /api -q 2>/dev/null; do
-          echo kubectl auth can-i get /api: exit status $?
-          sleep 2
-        done
-      '';
       serviceConfig = {
         RestartSec = "30s";
         Restart = "on-failure";
@@ -140,7 +120,7 @@ in
             "--cluster-cidr=${cfg.clusterCidr}"} \
           ${optionalString (cfg.featureGates != [])
             "--feature-gates=${concatMapStringsSep "," (feature: "${feature}=true") cfg.featureGates}"} \
-          --kubeconfig=${environment.KUBECONFIG} \
+          --kubeconfig=${top.lib.mkKubeConfig "kube-controller-manager" cfg.kubeconfig} \
           --leader-elect=${boolToString cfg.leaderElect} \
           ${optionalString (cfg.rootCaFile!=null)
             "--root-ca-file=${cfg.rootCaFile}"} \
@@ -161,16 +141,7 @@ in
         User = "kubernetes";
         Group = "kubernetes";
       };
-      path = top.path ++ [ pkgs.kubectl ];
-      unitConfig.ConditionPathExists = controllerManagerPaths;
-    };
-
-    systemd.paths.kube-controller-manager = {
-      wantedBy = [ "kube-controller-manager.service" ];
-      pathConfig = {
-        PathExists = controllerManagerPaths;
-        PathChanged = controllerManagerPaths;
-      };
+      path = top.path;
     };
 
     services.kubernetes.pki.certs = with top.lib; {
diff --git a/nixos/modules/services/cluster/kubernetes/default.nix b/nixos/modules/services/cluster/kubernetes/default.nix
index 143b41f57f6a..3790ac9b6918 100644
--- a/nixos/modules/services/cluster/kubernetes/default.nix
+++ b/nixos/modules/services/cluster/kubernetes/default.nix
@@ -256,29 +256,6 @@ in {
         wantedBy = [ "multi-user.target" ];
       };
 
-      systemd.targets.kube-control-plane-online = {
-        wantedBy = [ "kubernetes.target" ];
-        before = [ "kubernetes.target" ];
-      };
-
-      systemd.services.kube-control-plane-online = rec {
-        description = "Kubernetes control plane is online";
-        wantedBy = [ "kube-control-plane-online.target" ];
-        after = [ "kube-scheduler.service" "kube-controller-manager.service" ];
-        before = [ "kube-control-plane-online.target" ];
-        path = [ pkgs.curl ];
-        preStart = ''
-          until curl -Ssf ${cfg.apiserverAddress}/healthz do
-            echo curl -Ssf ${cfg.apiserverAddress}/healthz: exit status $?
-            sleep 3
-          done
-        '';
-        script = "echo Ok";
-        serviceConfig = {
-          TimeoutSec = "500";
-        };
-      };
-
       systemd.tmpfiles.rules = [
         "d /opt/cni/bin 0755 root root -"
         "d /run/kubernetes 0755 kubernetes kubernetes -"
@@ -302,8 +279,6 @@ in {
       services.kubernetes.apiserverAddress = mkDefault ("https://${if cfg.apiserver.advertiseAddress != null
                           then cfg.apiserver.advertiseAddress
                           else "${cfg.masterAddress}:${toString cfg.apiserver.securePort}"}");
-
-      services.kubernetes.kubeconfig.server = mkDefault cfg.apiserverAddress;
     })
   ];
 }
diff --git a/nixos/modules/services/cluster/kubernetes/flannel.nix b/nixos/modules/services/cluster/kubernetes/flannel.nix
index d9437427d6d1..93ee2fd65eeb 100644
--- a/nixos/modules/services/cluster/kubernetes/flannel.nix
+++ b/nixos/modules/services/cluster/kubernetes/flannel.nix
@@ -23,27 +23,17 @@ in
 {
   ###### interface
   options.services.kubernetes.flannel = {
-    enable = mkEnableOption "flannel networking";
-    kubeconfig = top.lib.mkKubeConfigOptions "Kubernetes flannel";
+    enable = mkEnableOption "enable flannel networking";
   };
 
   ###### implementation
-  config = let
-
-    flannelPaths = filter (a: a != null) [
-      cfg.kubeconfig.caFile
-      cfg.kubeconfig.certFile
-      cfg.kubeconfig.keyFile
-    ];
-    kubeconfig = top.lib.mkKubeConfig "flannel" cfg.kubeconfig;
-
-  in mkIf cfg.enable {
+  config = mkIf cfg.enable {
     services.flannel = {
 
       enable = mkDefault true;
       network = mkDefault top.clusterCidr;
-      inherit storageBackend kubeconfig;
-      nodeName = top.kubelet.hostname;
+      inherit storageBackend;
+      nodeName = config.services.kubernetes.kubelet.hostname;
     };
 
     services.kubernetes.kubelet = {
@@ -58,66 +48,24 @@ in
       }];
     };
 
-    systemd.services.mk-docker-opts = {
+    systemd.services."mk-docker-opts" = {
       description = "Pre-Docker Actions";
-      wantedBy = [ "flannel.target" ];
-      before = [ "flannel.target" ];
       path = with pkgs; [ gawk gnugrep ];
       script = ''
         ${mkDockerOpts}/mk-docker-opts -d /run/flannel/docker
         systemctl restart docker
       '';
-      unitConfig.ConditionPathExists = [ "/run/flannel/subnet.env" ];
       serviceConfig.Type = "oneshot";
     };
 
-    systemd.paths.flannel-subnet-env = {
-      wantedBy = [ "mk-docker-opts.service" ];
+    systemd.paths."flannel-subnet-env" = {
+      wantedBy = [ "flannel.service" ];
       pathConfig = {
-        PathExists = [ "/run/flannel/subnet.env" ];
-        PathChanged = [ "/run/flannel/subnet.env" ];
+        PathModified = "/run/flannel/subnet.env";
         Unit = "mk-docker-opts.service";
       };
     };
 
-    systemd.targets.flannel = {
-      wantedBy = [ "kube-node-online.target" ];
-      before = [ "kube-node-online.target" ];
-    };
-
-    systemd.services.flannel = {
-      wantedBy = [ "flannel.target" ];
-      after = [ "kubelet.target" ];
-      before = [ "flannel.target" ];
-      path = with pkgs; [ iptables kubectl ];
-      environment.KUBECONFIG = kubeconfig;
-      preStart = let
-        args = [
-          "--selector=kubernetes.io/hostname=${top.kubelet.hostname}"
-          # flannel exits if node is not registered yet, before that there is no podCIDR
-          "--output=jsonpath={.items[0].spec.podCIDR}"
-          # if jsonpath cannot be resolved exit with status 1
-          "--allow-missing-template-keys=false"
-        ];
-      in ''
-        until kubectl get nodes ${concatStringsSep " " args} 2>/dev/null; do
-          echo Waiting for ${top.kubelet.hostname} to be RegisteredNode
-          sleep 1
-        done
-      '';
-      unitConfig.ConditionPathExists = flannelPaths;
-    };
-
-    systemd.paths.flannel = {
-      wantedBy = [ "flannel.service" ];
-      pathConfig = {
-        PathExists = flannelPaths;
-        PathChanged = flannelPaths;
-      };
-    };
-
-    services.kubernetes.flannel.kubeconfig.server = mkDefault top.apiserverAddress;
-
     systemd.services.docker = {
       environment.DOCKER_OPTS = "-b none";
       serviceConfig.EnvironmentFile = "-/run/flannel/docker";
@@ -144,6 +92,7 @@ in
 
     # give flannel som kubernetes rbac permissions if applicable
     services.kubernetes.addonManager.bootstrapAddons = mkIf ((storageBackend == "kubernetes") && (elem "RBAC" top.apiserver.authorizationMode)) {
+
       flannel-cr = {
         apiVersion = "rbac.authorization.k8s.io/v1beta1";
         kind = "ClusterRole";
@@ -179,6 +128,7 @@ in
           name = "flannel-client";
         }];
       };
+
     };
   };
 }
diff --git a/nixos/modules/services/cluster/kubernetes/kubelet.nix b/nixos/modules/services/cluster/kubernetes/kubelet.nix
index 4c5df96bcc6a..e93062d97437 100644
--- a/nixos/modules/services/cluster/kubernetes/kubelet.nix
+++ b/nixos/modules/services/cluster/kubernetes/kubelet.nix
@@ -234,28 +234,21 @@ in
 
   ###### implementation
   config = mkMerge [
-    (let
-
-      kubeletPaths = filter (a: a != null) [
-        cfg.kubeconfig.caFile
-        cfg.kubeconfig.certFile
-        cfg.kubeconfig.keyFile
-        cfg.clientCaFile
-        cfg.tlsCertFile
-        cfg.tlsKeyFile
-      ];
-
-    in mkIf cfg.enable {
+    (mkIf cfg.enable {
       services.kubernetes.kubelet.seedDockerImages = [infraContainer];
 
       systemd.services.kubelet = {
         description = "Kubernetes Kubelet Service";
-        wantedBy = [ "kubelet.target" ];
-        after = [ "kube-control-plane-online.target" ];
-        before = [ "kubelet.target" ];
+        wantedBy = [ "kubernetes.target" ];
+        after = [ "network.target" "docker.service" "kube-apiserver.service" ];
         path = with pkgs; [ gitMinimal openssh docker utillinux iproute ethtool thin-provisioning-tools iptables socat ] ++ top.path;
         preStart = ''
-          rm -f /opt/cni/bin/* || true
+          ${concatMapStrings (img: ''
+            echo "Seeding docker image: ${img}"
+            docker load <${img}
+          '') cfg.seedDockerImages}
+
+          rm /opt/cni/bin/* || true
           ${concatMapStrings (package: ''
             echo "Linking cni package: ${package}"
             ln -fs ${package}/bin/* /opt/cni/bin
@@ -308,56 +301,6 @@ in
           '';
           WorkingDirectory = top.dataDir;
         };
-        unitConfig.ConditionPathExists = kubeletPaths;
-      };
-
-      systemd.paths.kubelet = {
-        wantedBy =  [ "kubelet.service" ];
-        pathConfig = {
-          PathExists = kubeletPaths;
-          PathChanged = kubeletPaths;
-        };
-      };
-
-      systemd.services.docker.before = [ "kubelet.service" ];
-
-      systemd.services.docker-seed-images = {
-        wantedBy = [ "docker.service" ];
-        after = [ "docker.service" ];
-        before = [ "kubelet.service" ];
-        path = with pkgs; [ docker ];
-        preStart = ''
-          ${concatMapStrings (img: ''
-            echo "Seeding docker image: ${img}"
-            docker load <${img}
-          '') cfg.seedDockerImages}
-        '';
-        script = "echo Ok";
-        serviceConfig.Type = "oneshot";
-        serviceConfig.RemainAfterExit = true;
-        serviceConfig.Slice = "kubernetes.slice";
-      };
-
-      systemd.services.kubelet-online = {
-        wantedBy = [ "kube-node-online.target" ];
-        after = [ "flannel.target" "kubelet.target" ];
-        before = [ "kube-node-online.target" ];
-        # it is complicated. flannel needs kubelet to run the pause container before
-        # it discusses the node CIDR with apiserver and afterwards configures and restarts
-        # dockerd. Until then prevent creating any pods because they have to be recreated anyway
-        # because the network of docker0 has been changed by flannel.
-        script = let
-          docker-env = "/run/flannel/docker";
-          flannel-date = "stat --print=%Y ${docker-env}";
-          docker-date = "systemctl show --property=ActiveEnterTimestamp --value docker";
-        in ''
-          until test -f ${docker-env} ; do sleep 1 ; done
-          while test `${flannel-date}` -gt `date +%s --date="$(${docker-date})"` ; do
-            sleep 1
-          done
-        '';
-        serviceConfig.Type = "oneshot";
-        serviceConfig.Slice = "kubernetes.slice";
       };
 
       # Allways include cni plugins
@@ -404,16 +347,5 @@ in
       };
     })
 
-    {
-      systemd.targets.kubelet = {
-        wantedBy = [ "kube-node-online.target" ];
-        before = [ "kube-node-online.target" ];
-      };
-
-      systemd.targets.kube-node-online = {
-        wantedBy = [ "kubernetes.target" ];
-        before = [ "kubernetes.target" ];
-      };
-    }
   ];
 }
diff --git a/nixos/modules/services/cluster/kubernetes/pki.nix b/nixos/modules/services/cluster/kubernetes/pki.nix
index 47384ae50a07..733479e24c97 100644
--- a/nixos/modules/services/cluster/kubernetes/pki.nix
+++ b/nixos/modules/services/cluster/kubernetes/pki.nix
@@ -27,11 +27,12 @@ let
   certmgrAPITokenPath = "${top.secretsPath}/${cfsslAPITokenBaseName}";
   cfsslAPITokenLength = 32;
 
-  clusterAdminKubeconfig = with cfg.certs.clusterAdmin; {
-    server = top.apiserverAddress;
-    certFile = cert;
-    keyFile = key;
-  };
+  clusterAdminKubeconfig = with cfg.certs.clusterAdmin;
+    top.lib.mkKubeConfig "cluster-admin" {
+        server = top.apiserverAddress;
+        certFile = cert;
+        keyFile = key;
+    };
 
   remote = with config.services; "https://${kubernetes.masterAddress}:${toString cfssl.port}";
 in
@@ -118,11 +119,6 @@ in
     cfsslCertPathPrefix = "${config.services.cfssl.dataDir}/cfssl";
     cfsslCert = "${cfsslCertPathPrefix}.pem";
     cfsslKey = "${cfsslCertPathPrefix}-key.pem";
-
-    certmgrPaths = [
-      top.caFile
-      certmgrAPITokenPath
-    ];
   in
   {
 
@@ -172,40 +168,13 @@ in
         chown cfssl "${cfsslAPITokenPath}" && chmod 400 "${cfsslAPITokenPath}"
       '')]);
 
-    systemd.targets.cfssl-online = {
-      wantedBy = [ "network-online.target" ];
-      after = [ "cfssl.service" "network-online.target" "cfssl-online.service" ];
-    };
-
-    systemd.services.cfssl-online = {
-      description = "Wait for ${remote} to be reachable.";
-      wantedBy = [ "cfssl-online.target" ];
-      before = [ "cfssl-online.target" ];
-      path = [ pkgs.curl ];
-      preStart = ''
-        until curl --fail-early -fskd '{}' ${remote}/api/v1/cfssl/info -o /dev/null; do
-          echo curl ${remote}/api/v1/cfssl/info: exit status $?
-          sleep 2
-        done
-      '';
-      script = "echo Ok";
-      serviceConfig = {
-        TimeoutSec = "300";
-      };
-    };
-
     systemd.services.kube-certmgr-bootstrap = {
       description = "Kubernetes certmgr bootstrapper";
-      wantedBy = [ "cfssl-online.target" ];
-      after = [ "cfssl-online.target" ];
-      before = [ "certmgr.service" ];
-      path = with pkgs; [ curl cfssl ];
+      wantedBy = [ "certmgr.service" ];
+      after = [ "cfssl.target" ];
       script = concatStringsSep "\n" [''
         set -e
 
-        mkdir -p $(dirname ${certmgrAPITokenPath})
-        mkdir -p $(dirname ${top.caFile})
-
         # If there's a cfssl (cert issuer) running locally, then don't rely on user to
         # manually paste it in place. Just symlink.
         # otherwise, create the target file, ready for users to insert the token
@@ -217,18 +186,15 @@ in
         fi
       ''
       (optionalString (cfg.pkiTrustOnBootstrap) ''
-        if [ ! -s "${top.caFile}" ]; then
-          until test -s ${top.caFile}.json; do
-            sleep 2
-            curl --fail-early -fskd '{}' ${remote}/api/v1/cfssl/info -o ${top.caFile}.json
-          done
-          cfssljson -f ${top.caFile}.json -stdout >${top.caFile}
-          rm ${top.caFile}.json
+        if [ ! -f "${top.caFile}" ] || [ $(cat "${top.caFile}" | wc -c) -lt 1 ]; then
+          ${pkgs.curl}/bin/curl --fail-early -f -kd '{}' ${remote}/api/v1/cfssl/info | \
+            ${pkgs.cfssl}/bin/cfssljson -stdout >${top.caFile}
         fi
       '')
       ];
       serviceConfig = {
-        TimeoutSec = "500";
+        RestartSec = "10s";
+        Restart = "on-failure";
       };
     };
 
@@ -264,28 +230,35 @@ in
           mapAttrs mkSpec cfg.certs;
       };
 
-      systemd.services.certmgr = {
-        wantedBy = [ "cfssl-online.target" ];
-        after = [ "cfssl-online.target" "kube-certmgr-bootstrap.service" ];
-        preStart = ''
-          while ! test -s ${certmgrAPITokenPath} ; do
-            sleep 1
-            echo Waiting for ${certmgrAPITokenPath}
-          done
-        '';
-        unitConfig.ConditionPathExists = certmgrPaths;
-      };
+      #TODO: Get rid of kube-addon-manager in the future for the following reasons
+      # - it is basically just a shell script wrapped around kubectl
+      # - it assumes that it is clusterAdmin or can gain clusterAdmin rights through serviceAccount
+      # - it is designed to be used with k8s system components only
+      # - it would be better with a more Nix-oriented way of managing addons
+      systemd.services.kube-addon-manager = mkIf top.addonManager.enable (mkMerge [{
+        environment.KUBECONFIG = with cfg.certs.addonManager;
+          top.lib.mkKubeConfig "addon-manager" {
+            server = top.apiserverAddress;
+            certFile = cert;
+            keyFile = key;
+          };
+        }
 
-      systemd.paths.certmgr = {
-        wantedBy = [ "certmgr.service" ];
-        pathConfig = {
-          PathExists = certmgrPaths;
-          PathChanged = certmgrPaths;
-        };
-      };
+        (optionalAttrs (top.addonManager.bootstrapAddons != {}) {
+          serviceConfig.PermissionsStartOnly = true;
+          preStart = with pkgs;
+          let
+            files = mapAttrsToList (n: v: writeText "${n}.json" (builtins.toJSON v))
+              top.addonManager.bootstrapAddons;
+          in
+          ''
+            export KUBECONFIG=${clusterAdminKubeconfig}
+            ${kubectl}/bin/kubectl apply -f ${concatStringsSep " \\\n -f " files}
+          '';
+        })]);
 
-      environment.etc.${cfg.etcClusterAdminKubeconfig}.source = mkIf (cfg.etcClusterAdminKubeconfig != null)
-        (top.lib.mkKubeConfig "cluster-admin" clusterAdminKubeconfig);
+      environment.etc.${cfg.etcClusterAdminKubeconfig}.source = mkIf (!isNull cfg.etcClusterAdminKubeconfig)
+        clusterAdminKubeconfig;
 
       environment.systemPackages = mkIf (top.kubelet.enable || top.proxy.enable) [
       (pkgs.writeScriptBin "nixos-kubernetes-node-join" ''
@@ -311,22 +284,38 @@ in
           exit 1
         fi
 
-        do_restart=$(test -s ${certmgrAPITokenPath} && echo -n y || echo -n n)
-
         echo $token > ${certmgrAPITokenPath}
         chmod 600 ${certmgrAPITokenPath}
 
-        if [ y = $do_restart ]; then
-          echo "Restarting certmgr..." >&1
-          systemctl restart certmgr
-        fi
+        echo "Restarting certmgr..." >&1
+        systemctl restart certmgr
 
-        echo "Node joined succesfully" >&1
+        echo "Waiting for certs to appear..." >&1
+
+        ${optionalString top.kubelet.enable ''
+          while [ ! -f ${cfg.certs.kubelet.cert} ]; do sleep 1; done
+          echo "Restarting kubelet..." >&1
+          systemctl restart kubelet
+        ''}
+
+        ${optionalString top.proxy.enable ''
+          while [ ! -f ${cfg.certs.kubeProxyClient.cert} ]; do sleep 1; done
+          echo "Restarting kube-proxy..." >&1
+          systemctl restart kube-proxy
+        ''}
+
+        ${optionalString top.flannel.enable ''
+          while [ ! -f ${cfg.certs.flannelClient.cert} ]; do sleep 1; done
+          echo "Restarting flannel..." >&1
+          systemctl restart flannel
+        ''}
+
+        echo "Node joined succesfully"
       '')];
 
       # isolate etcd on loopback at the master node
       # easyCerts doesn't support multimaster clusters anyway atm.
-      services.etcd = mkIf top.apiserver.enable (with cfg.certs.etcd; {
+      services.etcd = with cfg.certs.etcd; {
         listenClientUrls = ["https://127.0.0.1:2379"];
         listenPeerUrls = ["https://127.0.0.1:2380"];
         advertiseClientUrls = ["https://etcd.local:2379"];
@@ -335,11 +324,19 @@ in
         certFile = mkDefault cert;
         keyFile = mkDefault key;
         trustedCaFile = mkDefault caCert;
-      });
+      };
       networking.extraHosts = mkIf (config.services.etcd.enable) ''
         127.0.0.1 etcd.${top.addons.dns.clusterDomain} etcd.local
       '';
 
+      services.flannel = with cfg.certs.flannelClient; {
+        kubeconfig = top.lib.mkKubeConfig "flannel" {
+          server = top.apiserverAddress;
+          certFile = cert;
+          keyFile = key;
+        };
+      };
+
       services.kubernetes = {
 
         apiserver = mkIf top.apiserver.enable (with cfg.certs.apiServer; {
@@ -359,13 +356,6 @@ in
           proxyClientCertFile = mkDefault cfg.certs.apiserverProxyClient.cert;
           proxyClientKeyFile = mkDefault cfg.certs.apiserverProxyClient.key;
         });
-        addonManager = mkIf top.addonManager.enable {
-          kubeconfig = with cfg.certs.addonManager; {
-            certFile = mkDefault cert;
-            keyFile = mkDefault key;
-          };
-          bootstrapAddonsKubeconfig = clusterAdminKubeconfig;
-        };
         controllerManager = mkIf top.controllerManager.enable {
           serviceAccountKeyFile = mkDefault cfg.certs.serviceAccount.key;
           rootCaFile = cfg.certs.controllerManagerClient.caCert;
@@ -374,12 +364,6 @@ in
             keyFile = mkDefault key;
           };
         };
-        flannel = mkIf top.flannel.enable {
-          kubeconfig = with cfg.certs.flannelClient; {
-            certFile = cert;
-            keyFile = key;
-          };
-        };
         scheduler = mkIf top.scheduler.enable {
           kubeconfig = with cfg.certs.schedulerClient; {
             certFile = mkDefault cert;
diff --git a/nixos/modules/services/cluster/kubernetes/proxy.nix b/nixos/modules/services/cluster/kubernetes/proxy.nix
index 23f4d97b7030..bd4bf04ea833 100644
--- a/nixos/modules/services/cluster/kubernetes/proxy.nix
+++ b/nixos/modules/services/cluster/kubernetes/proxy.nix
@@ -45,28 +45,12 @@ in
   };
 
   ###### implementation
-  config = let
-
-    proxyPaths = filter (a: a != null) [
-      cfg.kubeconfig.caFile
-      cfg.kubeconfig.certFile
-      cfg.kubeconfig.keyFile
-    ];
-
-  in mkIf cfg.enable {
-    systemd.services.kube-proxy = rec {
+  config = mkIf cfg.enable {
+    systemd.services.kube-proxy = {
       description = "Kubernetes Proxy Service";
-      wantedBy = [ "kube-node-online.target" ];
-      after = [ "kubelet-online.service" ];
-      before = [ "kube-node-online.target" ];
-      environment.KUBECONFIG = top.lib.mkKubeConfig "kube-proxy" cfg.kubeconfig;
-      path = with pkgs; [ iptables conntrack_tools kubectl ];
-      preStart = ''
-        until kubectl auth can-i get nodes/${top.kubelet.hostname} -q 2>/dev/null; do
-          echo kubectl auth can-i get nodes/${top.kubelet.hostname}: exit status $?
-          sleep 2
-        done
-      '';
+      wantedBy = [ "kubernetes.target" ];
+      after = [ "kube-apiserver.service" ];
+      path = with pkgs; [ iptables conntrack_tools ];
       serviceConfig = {
         Slice = "kubernetes.slice";
         ExecStart = ''${top.package}/bin/kube-proxy \
@@ -75,7 +59,7 @@ in
             "--cluster-cidr=${top.clusterCidr}"} \
           ${optionalString (cfg.featureGates != [])
             "--feature-gates=${concatMapStringsSep "," (feature: "${feature}=true") cfg.featureGates}"} \
-          --kubeconfig=${environment.KUBECONFIG} \
+          --kubeconfig=${top.lib.mkKubeConfig "kube-proxy" cfg.kubeconfig} \
           ${optionalString (cfg.verbosity != null) "--v=${toString cfg.verbosity}"} \
           ${cfg.extraOpts}
         '';
@@ -83,15 +67,6 @@ in
         Restart = "on-failure";
         RestartSec = 5;
       };
-      unitConfig.ConditionPathExists = proxyPaths;
-    };
-
-    systemd.paths.kube-proxy = {
-      wantedBy = [ "kube-proxy.service" ];
-      pathConfig = {
-        PathExists = proxyPaths;
-        PathChanged = proxyPaths;
-      };
     };
 
     services.kubernetes.pki.certs = {
diff --git a/nixos/modules/services/cluster/kubernetes/scheduler.nix b/nixos/modules/services/cluster/kubernetes/scheduler.nix
index a0e484542951..5f6113227d9d 100644
--- a/nixos/modules/services/cluster/kubernetes/scheduler.nix
+++ b/nixos/modules/services/cluster/kubernetes/scheduler.nix
@@ -56,35 +56,18 @@ in
   };
 
   ###### implementation
-  config =  let
-
-    schedulerPaths = filter (a: a != null) [
-      cfg.kubeconfig.caFile
-      cfg.kubeconfig.certFile
-      cfg.kubeconfig.keyFile
-    ];
-
-  in mkIf cfg.enable {
-    systemd.services.kube-scheduler = rec {
+  config = mkIf cfg.enable {
+    systemd.services.kube-scheduler = {
       description = "Kubernetes Scheduler Service";
-      wantedBy = [ "kube-control-plane-online.target" ];
+      wantedBy = [ "kubernetes.target" ];
       after = [ "kube-apiserver.service" ];
-      before = [ "kube-control-plane-online.target" ];
-      environment.KUBECONFIG = top.lib.mkKubeConfig "kube-scheduler" cfg.kubeconfig;
-      path = [ pkgs.kubectl ];
-      preStart = ''
-        until kubectl auth can-i get /api -q 2>/dev/null; do
-          echo kubectl auth can-i get /api: exit status $?
-          sleep 2
-        done
-      '';
       serviceConfig = {
         Slice = "kubernetes.slice";
         ExecStart = ''${top.package}/bin/kube-scheduler \
           --address=${cfg.address} \
           ${optionalString (cfg.featureGates != [])
             "--feature-gates=${concatMapStringsSep "," (feature: "${feature}=true") cfg.featureGates}"} \
-          --kubeconfig=${environment.KUBECONFIG} \
+          --kubeconfig=${top.lib.mkKubeConfig "kube-scheduler" cfg.kubeconfig} \
           --leader-elect=${boolToString cfg.leaderElect} \
           --port=${toString cfg.port} \
           ${optionalString (cfg.verbosity != null) "--v=${toString cfg.verbosity}"} \
@@ -96,15 +79,6 @@ in
         Restart = "on-failure";
         RestartSec = 5;
       };
-      unitConfig.ConditionPathExists = schedulerPaths;
-    };
-
-    systemd.paths.kube-scheduler = {
-      wantedBy = [ "kube-scheduler.service" ];
-      pathConfig = {
-        PathExists = schedulerPaths;
-        PathChanged = schedulerPaths;
-      };
     };
 
     services.kubernetes.pki.certs = {
diff --git a/nixos/tests/kubernetes/base.nix b/nixos/tests/kubernetes/base.nix
index 212023859f6d..ec1a75e74c41 100644
--- a/nixos/tests/kubernetes/base.nix
+++ b/nixos/tests/kubernetes/base.nix
@@ -30,10 +30,7 @@ let
         { config, pkgs, lib, nodes, ... }:
           mkMerge [
             {
-              boot = {
-                postBootCommands = "rm -fr /var/lib/kubernetes/secrets /tmp/shared/*";
-                kernel.sysctl = { "fs.inotify.max_user_instances" = 256; };
-              };
+              boot.postBootCommands = "rm -fr /var/lib/kubernetes/secrets /tmp/shared/*";
               virtualisation.memorySize = mkDefault 1536;
               virtualisation.diskSize = mkDefault 4096;
               networking = {
diff --git a/nixos/tests/kubernetes/dns.nix b/nixos/tests/kubernetes/dns.nix
index e7db0a58ab61..46bcb01a5265 100644
--- a/nixos/tests/kubernetes/dns.nix
+++ b/nixos/tests/kubernetes/dns.nix
@@ -77,7 +77,6 @@ let
   singleNodeTest = {
     test = ''
       # prepare machine1 for test
-      $machine1->waitForUnit("kubernetes.target");
       $machine1->waitUntilSucceeds("kubectl get node machine1.${domain} | grep -w Ready");
       $machine1->waitUntilSucceeds("docker load < ${redisImage}");
       $machine1->waitUntilSucceeds("kubectl create -f ${redisPod}");
@@ -103,8 +102,6 @@ let
       # Node token exchange
       $machine1->waitUntilSucceeds("cp -f /var/lib/cfssl/apitoken.secret /tmp/shared/apitoken.secret");
       $machine2->waitUntilSucceeds("cat /tmp/shared/apitoken.secret | nixos-kubernetes-node-join");
-      $machine1->waitForUnit("kubernetes.target");
-      $machine2->waitForUnit("kubernetes.target");
 
       # prepare machines for test
       $machine1->waitUntilSucceeds("kubectl get node machine2.${domain} | grep -w Ready");
diff --git a/nixos/tests/kubernetes/rbac.nix b/nixos/tests/kubernetes/rbac.nix
index 967fe506004f..3ce7adcd0d71 100644
--- a/nixos/tests/kubernetes/rbac.nix
+++ b/nixos/tests/kubernetes/rbac.nix
@@ -94,8 +94,6 @@ let
 
   singlenode = base // {
     test = ''
-      $machine1->waitForUnit("kubernetes.target");
-
       $machine1->waitUntilSucceeds("kubectl get node machine1.my.zyx | grep -w Ready");
 
       $machine1->waitUntilSucceeds("docker load < ${kubectlImage}");
@@ -118,8 +116,6 @@ let
       # Node token exchange
       $machine1->waitUntilSucceeds("cp -f /var/lib/cfssl/apitoken.secret /tmp/shared/apitoken.secret");
       $machine2->waitUntilSucceeds("cat /tmp/shared/apitoken.secret | nixos-kubernetes-node-join");
-      $machine1->waitForUnit("kubernetes.target");
-      $machine2->waitForUnit("kubernetes.target");
 
       $machine1->waitUntilSucceeds("kubectl get node machine2.my.zyx | grep -w Ready");