Merge pull request #26828 from vishh/oom-kill-enable

Automatic merge from submit-queue

Enable support for memory eviction configuration via salt

Added evictions based on memory by default whenever the available memory is < 100Mi.

Updated GCE and GCI.

cluster/common.sh

@@ -647,6 +647,11 @@ EOF
   if [ -n "${NODE_LABELS:-}" ]; then
       cat >>$file <<EOF
 NODE_LABELS: $(yaml-quote ${NODE_LABELS})
+EOF
+    fi
+  if [ -n "${EVICTION_HARD:-}" ]; then
+      cat >>$file <<EOF
+EVICTION_HARD: $(yaml-quote ${EVICTION_HARD})
 EOF
     fi
   if [[ "${OS_DISTRIBUTION}" == "coreos" ]]; then

cluster/gce/config-default.sh

@@ -138,3 +138,6 @@ OPENCONTRAIL_PUBLIC_SUBNET="${OPENCONTRAIL_PUBLIC_SUBNET:-10.1.0.0/16}"
 HAIRPIN_MODE="${HAIRPIN_MODE:-promiscuous-bridge}" # promiscuous-bridge, hairpin-veth, none
 # Optional: if set to true, kube-up will configure the cluster to run e2e tests.
 E2E_STORAGE_TEST_ENVIRONMENT=${KUBE_E2E_STORAGE_TEST_ENVIRONMENT:-false}
+
+# Evict pods whenever compute resource availability on the nodes gets below a threshold.
+EVICTION_HARD="${EVICTION_HARD:-memory.available<100Mi}"

cluster/gce/config-test.sh

@@ -170,3 +170,6 @@ E2E_STORAGE_TEST_ENVIRONMENT=${KUBE_E2E_STORAGE_TEST_ENVIRONMENT:-false}
 # Optional: if set to true, a image puller is deployed. Only for use in e2e clusters.
 # TODO: Pipe this through GKE e2e clusters once we know it helps.
 PREPULL_E2E_IMAGES="${PREPULL_E2E_IMAGES:-true}"
+
+# Evict pods whenever compute resource availability on the nodes gets below a threshold.
+EVICTION_HARD="${EVICTION_HARD:-memory.available<100Mi}"

cluster/gce/configure-vm.sh

@@ -535,6 +535,11 @@ EOF
     if [ -n "${NODE_LABELS:-}" ]; then
       cat <<EOF >>/srv/salt-overlay/pillar/cluster-params.sls
 node_labels: '$(echo "${NODE_LABELS}" | sed -e "s/'/''/g")'
+EOF
+    fi
+    if [ -n "${EVICTION_HARD:-}" ]; then
+      cat <<EOF >>/srv/salt-overlay/pillar/cluster-params.sls
+eviction_hard: '$(echo "${EVICTION_HARD}" | sed -e "s/'/''/g")'
 EOF
     fi
     if [[ "${ENABLE_NODE_AUTOSCALER:-false}" == "true" ]]; then

cluster/gce/gci/configure-helper.sh

@@ -351,6 +351,9 @@ function start-kubelet {
   if [[ -n "${NODE_LABELS:-}" ]]; then
     flags+=" --node-labels=${NODE_LABELS}"
   fi
+  if [[ -n "${EVICTION_HARD:-}" ]]; then
+    flags+=" --eviction-hard=${EVICTION_HARD}"
+  fi
   if [[ "${ALLOCATE_NODE_CIDRS:-}" == "true" ]]; then
      flags+=" --configure-cbr0=${ALLOCATE_NODE_CIDRS}"
   fi

cluster/saltbase/salt/kubelet/default

@@ -196,5 +196,10 @@
   {% set node_labels="--node-labels=" + pillar['node_labels'] %}
 {% endif -%}
 
+{% set eviction_hard = "" %}
+{% if pillar['eviction_hard'] is defined -%}
+  {% set eviction_hard="--eviction-hard=" + pillar['eviction_hard'] %}
+{% endif -%}
+
 # test_args has to be kept at the end, so they'll overwrite any prior configuration
-DAEMON_ARGS="{{daemon_args}} {{api_servers_with_port}} {{debugging_handlers}} {{hostname_override}} {{cloud_provider}} {{cloud_config}} {{config}} {{manifest_url}} --allow-privileged={{pillar['allow_privileged']}} {{log_level}} {{cluster_dns}} {{cluster_domain}} {{docker_root}} {{kubelet_root}} {{configure_cbr0}} {{non_masquerade_cidr}} {{cgroup_root}} {{system_container}} {{pod_cidr}} {{ master_kubelet_args }} {{cpu_cfs_quota}} {{network_plugin}} {{kubelet_port}} {{experimental_flannel_overlay}} {{ reconcile_cidr_args }} {{ hairpin_mode }} {{enable_custom_metrics}} {{runtime_container}} {{kubelet_container}} {{node_labels}} {{babysit_daemons}} {{test_args}}"
+DAEMON_ARGS="{{daemon_args}} {{api_servers_with_port}} {{debugging_handlers}} {{hostname_override}} {{cloud_provider}} {{cloud_config}} {{config}} {{manifest_url}} --allow-privileged={{pillar['allow_privileged']}} {{log_level}} {{cluster_dns}} {{cluster_domain}} {{docker_root}} {{kubelet_root}} {{configure_cbr0}} {{non_masquerade_cidr}} {{cgroup_root}} {{system_container}} {{pod_cidr}} {{ master_kubelet_args }} {{cpu_cfs_quota}} {{network_plugin}} {{kubelet_port}} {{experimental_flannel_overlay}} {{ reconcile_cidr_args }} {{ hairpin_mode }} {{enable_custom_metrics}} {{runtime_container}} {{kubelet_container}} {{node_labels}} {{babysit_daemons}} {{eviction_hard}} {{test_args}}"

hack/verify-flags-underscore.py

@@ -55,6 +55,8 @@ def get_all_files(rootdir):
         # don't visit certain dirs
         if 'vendor' in dirs:
             dirs.remove('vendor')
+        if '_output' in dirs:
+            dirs.remove('_output')
         if '_gopath' in dirs:
             dirs.remove('_gopath')
         if 'third_party' in dirs:

hack/verify-flags/exceptions.txt

@@ -8,8 +8,6 @@ cluster/aws/templates/configure-vm-aws.sh:  env-to-grains "hostname_override"
 cluster/aws/templates/configure-vm-aws.sh:  env-to-grains "runtime_config"
 cluster/aws/templates/configure-vm-aws.sh:  kubelet_api_servers: '${KUBELET_APISERVER}'
 cluster/centos/util.sh:  local node_ip=${node#*@}
-cluster/gce/config-default.sh:# reloads <os_distro>/helper.sh in the gap between when the master is created
-cluster/gce/config-test.sh:# reloads <os_distro>/helper.sh in the gap between when the master is created
 cluster/gce/configure-vm.sh:  advertise_address: '${EXTERNAL_IP}'
 cluster/gce/configure-vm.sh:  api_servers: '${KUBERNETES_MASTER_NAME}'
 cluster/gce/configure-vm.sh:  cloud_config: ${CLOUD_CONFIG}
@@ -46,6 +44,7 @@ cluster/saltbase/salt/kube-proxy/kube-proxy.manifest:  {% set cluster_cidr=" --c
 cluster/saltbase/salt/kubelet/default:  {% set api_servers_with_port = api_servers + ":6443" -%}
 cluster/saltbase/salt/kubelet/default:  {% set api_servers_with_port = api_servers -%}
 cluster/saltbase/salt/kubelet/default:  {% set enable_custom_metrics="--enable-custom-metrics=" + pillar['enable_custom_metrics'] %}
+cluster/saltbase/salt/kubelet/default:  {% set eviction_hard="--eviction-hard=" + pillar['eviction_hard'] %}
 cluster/saltbase/salt/kubelet/default:  {% set kubelet_port="--port=" + pillar['kubelet_port'] %}
 cluster/saltbase/salt/kubelet/default:  {% set node_labels="--node-labels=" + pillar['node_labels'] %}
 cluster/saltbase/salt/kubelet/default:{% if pillar.get('non_masquerade_cidr','') -%}