...

Text file src/k8s.io/kubernetes/cluster/validate-cluster.sh

Documentation: k8s.io/kubernetes/cluster

     1#!/usr/bin/env bash
     2
     3# Copyright 2014 The Kubernetes Authors.
     4#
     5# Licensed under the Apache License, Version 2.0 (the "License");
     6# you may not use this file except in compliance with the License.
     7# You may obtain a copy of the License at
     8#
     9#     http://www.apache.org/licenses/LICENSE-2.0
    10#
    11# Unless required by applicable law or agreed to in writing, software
    12# distributed under the License is distributed on an "AS IS" BASIS,
    13# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    14# See the License for the specific language governing permissions and
    15# limitations under the License.
    16
    17# Validates that the cluster is healthy.
    18# Error codes are:
    19# 0 - success
    20# 1 - fatal (cluster is unlikely to work)
    21# 2 - non-fatal (encountered some errors, but cluster should be working correctly)
    22
    23set -o errexit
    24set -o nounset
    25set -o pipefail
    26
    27KUBE_ROOT=$(dirname "${BASH_SOURCE[0]}")/..
    28
    29if [ -f "${KUBE_ROOT}/cluster/env.sh" ]; then
    30  source "${KUBE_ROOT}/cluster/env.sh"
    31fi
    32
    33source "${KUBE_ROOT}/hack/lib/util.sh"
    34source "${KUBE_ROOT}/cluster/kube-util.sh"
    35
    36# Run kubectl and retry upon failure.
    37function kubectl_retry() {
    38  tries=3
    39  while ! "${KUBE_ROOT}/cluster/kubectl.sh" "$@"; do
    40    tries=$((tries-1))
    41    if [[ ${tries} -le 0 ]]; then
    42      echo "('kubectl $*' failed, giving up)" >&2
    43      return 1
    44    fi
    45    echo "(kubectl failed, will retry ${tries} times)" >&2
    46    sleep 1
    47  done
    48}
    49
    50ALLOWED_NOTREADY_NODES="${ALLOWED_NOTREADY_NODES:-0}"
    51CLUSTER_READY_ADDITIONAL_TIME_SECONDS="${CLUSTER_READY_ADDITIONAL_TIME_SECONDS:-30}"
    52
    53if [[ "${KUBERNETES_PROVIDER:-}" == "gce" ]]; then
    54  if [[ "${KUBE_CREATE_NODES}" == "true" ]]; then
    55    EXPECTED_NUM_NODES="$(get-num-nodes)"
    56  else
    57    EXPECTED_NUM_NODES="0"
    58  fi
    59  echo "Validating gce cluster, MULTIZONE=${MULTIZONE:-}"
    60  # In multizone mode we need to add instances for all nodes in the region.
    61  if [[ "${MULTIZONE:-}" == "true" ]]; then
    62    EXPECTED_NUM_NODES=$(gcloud -q compute instances list --project="${PROJECT}" --format="[no-heading]" \
    63      --filter="(name ~ '${NODE_INSTANCE_PREFIX}.*' OR name ~ '${WINDOWS_NODE_INSTANCE_PREFIX}.*') AND zone:($(gcloud -q compute zones list --project="${PROJECT}" --filter=region="${REGION}" --format="csv[no-heading](name)" | tr "\n" "," | sed  "s/,$//"))" | wc -l)
    64    echo "Computing number of nodes, NODE_INSTANCE_PREFIX=${NODE_INSTANCE_PREFIX}, REGION=${REGION}, EXPECTED_NUM_NODES=${EXPECTED_NUM_NODES}"
    65  fi
    66else
    67  EXPECTED_NUM_NODES="${NUM_NODES}"
    68fi
    69
    70if [[ "${REGISTER_MASTER_KUBELET:-}" == "true" ]]; then
    71  if [[ "${KUBERNETES_PROVIDER:-}" == "gce" ]]; then
    72    NUM_MASTERS=$(get-master-replicas-count)
    73  else
    74    NUM_MASTERS=1
    75  fi
    76  EXPECTED_NUM_NODES=$((EXPECTED_NUM_NODES+NUM_MASTERS))
    77fi
    78
    79REQUIRED_NUM_NODES=$((EXPECTED_NUM_NODES - ALLOWED_NOTREADY_NODES))
    80# Make several attempts to deal with slow cluster birth.
    81return_value=0
    82attempt=0
    83# Set the timeout to ~25minutes (100 x 15 second) to avoid timeouts for 1000-node clusters.
    84PAUSE_BETWEEN_ITERATIONS_SECONDS=15
    85MAX_ATTEMPTS=100
    86ADDITIONAL_ITERATIONS=$(((CLUSTER_READY_ADDITIONAL_TIME_SECONDS + PAUSE_BETWEEN_ITERATIONS_SECONDS - 1)/PAUSE_BETWEEN_ITERATIONS_SECONDS))
    87while true; do
    88  # Pause between iterations of this large outer loop.
    89  if [[ ${attempt} -gt 0 ]]; then
    90    sleep 15
    91  fi
    92  attempt=$((attempt+1))
    93
    94  # The "kubectl get nodes -o template" exports node information.
    95  #
    96  # Echo the output and gather 2 counts:
    97  #  - Total number of nodes.
    98  #  - Number of "ready" nodes.
    99  #
   100  # Suppress errors from kubectl output because during cluster bootstrapping
   101  # for clusters where the master node is registered, the apiserver will become
   102  # available and then get restarted as the kubelet configures the docker bridge.
   103  #
   104  # We are assigning the result of kubectl_retry get nodes operation to the res
   105  # variable in that way, to prevent stopping the whole script on an error.
   106  #
   107  # Bash command substitution $(kubectl_...) removes all trailing whitespaces
   108  # which are important for line counting.
   109  # Use trick from https://unix.stackexchange.com/a/383411 to avoid
   110  # newline truncation.
   111  node=$(kubectl_retry get nodes --no-headers; ret=$?; echo .; exit "$ret") && res="$?" || res="$?"
   112  node="${node%.}"
   113  if [ "${res}" -ne "0" ]; then
   114    if [[ "${attempt}" -gt "${last_run:-$MAX_ATTEMPTS}" ]]; then
   115      echo -e "${color_red:-} Failed to get nodes.${color_norm:-}"
   116      exit 1
   117    else
   118      continue
   119    fi
   120  fi
   121  found=$(echo -n "${node}" | wc -l)
   122  # Use grep || true so that empty result doesn't return nonzero exit code.
   123  ready=$(echo -n "${node}" | grep -c -v "NotReady" || true)
   124
   125  if (( "${found}" == "${EXPECTED_NUM_NODES}" )) && (( "${ready}" == "${EXPECTED_NUM_NODES}")); then
   126    break
   127  elif (( "${found}" > "${EXPECTED_NUM_NODES}" )); then
   128    if [[ "${KUBE_USE_EXISTING_MASTER:-}" != "true" ]]; then
   129      echo -e "${color_red}Found ${found} nodes, but expected ${EXPECTED_NUM_NODES}. Your cluster may not behave correctly.${color_norm}"
   130    fi
   131    break
   132  elif (( "${ready}" > "${EXPECTED_NUM_NODES}")); then
   133    echo -e "${color_red}Found ${ready} ready nodes, but expected ${EXPECTED_NUM_NODES}. Your cluster may not behave correctly.${color_norm}"
   134    break
   135  else
   136    if [[ "${REQUIRED_NUM_NODES}" -le "${ready}" ]]; then
   137      echo -e "${color_green:-}Found ${REQUIRED_NUM_NODES} Nodes, allowing additional ${ADDITIONAL_ITERATIONS} iterations for other Nodes to join.${color_norm}"
   138      last_run="${last_run:-$((attempt + ADDITIONAL_ITERATIONS - 1))}"
   139    fi
   140    if [[ "${attempt}" -gt "${last_run:-$MAX_ATTEMPTS}" ]]; then
   141      echo -e "${color_yellow:-}Detected ${ready} ready nodes, found ${found} nodes out of expected ${EXPECTED_NUM_NODES}. Your cluster may not be fully functional.${color_norm}"
   142      kubectl_retry get nodes
   143      if [[ "${REQUIRED_NUM_NODES}" -gt "${ready}" ]]; then
   144        exit 1
   145      else
   146        return_value=2
   147        break
   148      fi
   149    else
   150      echo -e "${color_yellow}Waiting for ${EXPECTED_NUM_NODES} ready nodes. ${ready} ready nodes, ${found} registered. Retrying.${color_norm}"
   151    fi
   152  fi
   153done
   154echo "Found ${found} node(s)."
   155kubectl_retry get nodes
   156
   157attempt=0
   158while true; do
   159  # The "kubectl componentstatuses -o template" exports components health information.
   160  #
   161  # Echo the output and gather 2 counts:
   162  #  - Total number of componentstatuses.
   163  #  - Number of "healthy" components.
   164  cs_status=$(kubectl_retry get componentstatuses -o template --template='{{range .items}}{{with index .conditions 0}}{{.type}}:{{.status}}{{end}}{{"\n"}}{{end}}') || true
   165  componentstatuses=$(echo "${cs_status}" | grep -c 'Healthy:') || true
   166  healthy=$(echo "${cs_status}" | grep -c 'Healthy:True') || true
   167
   168  if ((componentstatuses > healthy)) || ((componentstatuses == 0)); then
   169    if ((attempt < 5)); then
   170      echo -e "${color_yellow}Cluster not working yet.${color_norm}"
   171      attempt=$((attempt+1))
   172      sleep 30
   173    else
   174      echo -e " ${color_yellow}Validate output:${color_norm}"
   175      kubectl_retry get cs
   176      echo -e "${color_red}Validation returned one or more failed components. Cluster is probably broken.${color_norm}"
   177      exit 1
   178    fi
   179  else
   180    break
   181  fi
   182done
   183
   184echo "Validate output:"
   185kubectl_retry get cs || true
   186if [ "${return_value}" == "0" ]; then
   187  echo -e "${color_green}Cluster validation succeeded${color_norm}"
   188else
   189  echo -e "${color_yellow}Cluster validation encountered some problems, but cluster should be in working order${color_norm}"
   190fi
   191
   192exit "${return_value}"

View as plain text