Salt configuration for Rescheduler

build/common.sh

@@ -944,6 +944,7 @@ function kube::release::package_kube_manifests_tarball() {
   cp "${salt_dir}/kube-controller-manager/kube-controller-manager.manifest" "${dst_dir}"
   cp "${salt_dir}/kube-addons/kube-addon-manager.yaml" "${dst_dir}"
   cp "${salt_dir}/l7-gcp/glbc.manifest" "${dst_dir}"
+  cp "${salt_dir}/rescheduler/rescheduler.manifest" "${dst_dir}/"
   cp "${KUBE_ROOT}/cluster/gce/trusty/configure-helper.sh" "${dst_dir}/trusty-configure-helper.sh"
   cp "${KUBE_ROOT}/cluster/gce/gci/configure-helper.sh" "${dst_dir}/gci-configure-helper.sh"
   cp "${KUBE_ROOT}/cluster/gce/gci/health-monitor.sh" "${dst_dir}/health-monitor.sh"

cluster/common.sh

@@ -550,6 +550,7 @@ ENABLE_CLUSTER_LOGGING: $(yaml-quote ${ENABLE_CLUSTER_LOGGING:-false})
 ENABLE_CLUSTER_UI: $(yaml-quote ${ENABLE_CLUSTER_UI:-false})
 ENABLE_NODE_PROBLEM_DETECTOR: $(yaml-quote ${ENABLE_NODE_PROBLEM_DETECTOR:-false})
 ENABLE_NODE_LOGGING: $(yaml-quote ${ENABLE_NODE_LOGGING:-false})
+ENABLE_RESCHEDULER: $(yaml-quote ${ENABLE_RESCHEDULER:-false})
 LOGGING_DESTINATION: $(yaml-quote ${LOGGING_DESTINATION:-})
 ELASTICSEARCH_LOGGING_REPLICAS: $(yaml-quote ${ELASTICSEARCH_LOGGING_REPLICAS:-})
 ENABLE_CLUSTER_DNS: $(yaml-quote ${ENABLE_CLUSTER_DNS:-false})

cluster/gce/config-default.sh

@@ -128,6 +128,9 @@ if [[ "${ENABLE_CLUSTER_AUTOSCALER}" == "true" ]]; then
   AUTOSCALER_ENABLE_SCALE_DOWN="${KUBE_AUTOSCALER_ENABLE_SCALE_DOWN:-true}"
 fi
 
+# Optional: Enable Rescheduler
+ENABLE_RESCHEDULER="${KUBE_ENABLE_RESCHEDULER:-false}"
+
 # Admission Controllers to invoke prior to persisting objects in cluster
 # If we included ResourceQuota, we should keep it at the end of the list to prevent incremeting quota usage prematurely.
 ADMISSION_CONTROL=NamespaceLifecycle,LimitRanger,ServiceAccount,PersistentVolumeLabel,ResourceQuota

cluster/gce/config-test.sh

@@ -148,6 +148,9 @@ if [[ "${ENABLE_CLUSTER_AUTOSCALER}" == "true" ]]; then
   AUTOSCALER_ENABLE_SCALE_DOWN="${KUBE_AUTOSCALER_ENABLE_SCALE_DOWN:-false}"
 fi
 
+# Optional: Enable Rescheduler
+ENABLE_RESCHEDULER="${KUBE_ENABLE_RESCHEDULER:-false}"
+
 # If we included ResourceQuota, we should keep it at the end of the list to prevent incremeting quota usage prematurely.
 ADMISSION_CONTROL="${KUBE_ADMISSION_CONTROL:-NamespaceLifecycle,LimitRanger,ServiceAccount,PersistentVolumeLabel,ResourceQuota}"
 

cluster/gce/configure-vm.sh

@@ -434,6 +434,7 @@ enable_cluster_ui: '$(echo "$ENABLE_CLUSTER_UI" | sed -e "s/'/''/g")'
 enable_node_problem_detector: '$(echo "$ENABLE_NODE_PROBLEM_DETECTOR" | sed -e "s/'/''/g")'
 enable_l7_loadbalancing: '$(echo "$ENABLE_L7_LOADBALANCING" | sed -e "s/'/''/g")'
 enable_node_logging: '$(echo "$ENABLE_NODE_LOGGING" | sed -e "s/'/''/g")'
+enable_rescheduler: '$(echo "$ENABLE_RESCHEDULER" | sed -e "s/'/''/g")'
 logging_destination: '$(echo "$LOGGING_DESTINATION" | sed -e "s/'/''/g")'
 elasticsearch_replicas: '$(echo "$ELASTICSEARCH_LOGGING_REPLICAS" | sed -e "s/'/''/g")'
 enable_cluster_dns: '$(echo "$ENABLE_CLUSTER_DNS" | sed -e "s/'/''/g")'

cluster/gce/gci/configure-helper.sh

@@ -978,6 +978,16 @@ function start-lb-controller {
   fi
 }
 
+# Starts rescheduler.
+function start-rescheduler {
+  if [[ "${ENABLE_RESCHEDULER:-}" == "true" ]]; then
+    echo "Starting Rescheduler"
+    prepare-log-file /var/log/rescheduler.log
+    cp "${KUBE_HOME}/kube-manifests/kubernetes/gci-trusty/rescheduler.manifest" \
+       /etc/kubernetes/manifests/
+  fi
+}
+
 function reset-motd {
   # kubelet is installed both on the master and nodes, and the version is easy to parse (unlike kubectl)
   local -r version="$(/usr/bin/kubelet --version=true | cut -f2 -d " ")"
@@ -1052,6 +1062,7 @@ if [[ "${KUBERNETES_MASTER:-}" == "true" ]]; then
   start-kube-addons
   start-cluster-autoscaler
   start-lb-controller
+  start-rescheduler
 else
   start-kube-proxy
   # Kube-registry-proxy.

cluster/gce/trusty/configure-helper.sh

@@ -701,6 +701,15 @@ start_cluster_autoscaler() {
   fi
 }
 
+# Starts rescheduler.
+start-rescheduler() {
+  if [[ "${ENABLE_RESCHEDULER:-}" == "true" ]]; then
+    prepare-log-file /var/log/rescheduler.log
+    cp "${KUBE_HOME}/kube-manifests/kubernetes/gci-trusty/rescheduler.manifest" \
+       /etc/kubernetes/manifests/
+  fi
+}
+
 # Starts a fluentd static pod for logging.
 start_fluentd() {
   if [ "${ENABLE_NODE_LOGGING:-}" = "true" ]; then

cluster/gce/trusty/master.yaml

@@ -192,6 +192,7 @@ script
 	start_kube_scheduler
 	start_kube_addons
 	start_cluster_autoscaler
+	start_rescheduler
 	reset_motd
 } 2>&1 | logger --priority daemon.info -t ${UPSTART_JOB}
 end script

cluster/saltbase/salt/rescheduler/init.sls

@@ -0,0 +1,15 @@
+/etc/kubernetes/manifests/rescheduler.manifest:
+  file.managed:
+    - source: salt://rescheduler/rescheduler.manifest
+    - template: jinja
+    - user: root
+    - group: root
+    - mode: 644
+    - makedirs: true
+    - dir_mode: 755
+
+/var/log/rescheduler.log:
+  file.managed:
+    - user: root
+    - group: root
+    - mode: 644

cluster/saltbase/salt/rescheduler/rescheduler.manifest

@@ -0,0 +1,35 @@
+apiVersion: v1
+kind: Pod
+metadata:
+  name: rescheduler-v0.1.0
+  namespace: kube-system
+  labels:
+    k8s-app: rescheduler
+    version: v0.1.0
+    kubernetes.io/cluster-service: "true"
+    kubernetes.io/name: "Rescheduler"
+spec:
+  hostNetwork: true
+  containers:
+  - image: gcr.io/google_containers/rescheduler:v0.1.0
+    name: rescheduler
+    volumeMounts:
+    - mountPath: /var/log/rescheduler.log
+      name: logfile
+      readOnly: false
+    resources:
+      limits:
+        cpu: 100m
+        memory: 300Mi
+      requests:
+        cpu: 10m
+        memory: 100Mi
+    command:
+    # TODO: split this out into args when we no longer need to pipe stdout to a file #6428
+    - sh
+    - -c
+    - '/rescheduler --running-in-cluster=false 1>>/var/log/rescheduler.log 2>&1'
+  volumes:
+  - hostPath:
+      path: /var/log/rescheduler.log
+    name: logfile

cluster/saltbase/salt/top.sls

@@ -101,6 +101,9 @@ base:
 {% if pillar.get('enable_cluster_autoscaler', '').lower() == 'true' %}
     - cluster-autoscaler
 {% endif %}
+{% if pillar.get('enable_rescheduler', '').lower() == 'true' %}
+    - rescheduler
+{% endif %}
 {% if pillar.get('network_policy_provider', '').lower() == 'calico' %}
     - calico.master
 {% endif %}