#! /bin/bash # Copyright 2016 The Kubernetes Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. set -eo pipefail # This script configures zookeeper cluster member ship for version of zookeeper # >= 3.5.0. It should not be used with the on-change.sh script in this example. # As of April-2016 is 3.4.8 is the latest stable. # Both /opt and /tmp/zookeeper are assumed to be volumes shared with the parent. # The format of each line in the dynamic config file is: # server.<1 based index>=::[:role];[:] # <1 based index> is the server index that matches the id in datadir/myid # is the port on which peers communicate to agree on updates # is the port used for leader election # [:role] can be set to observer, participant by default # is optional and defaults to 0.0.0.0 # is the port on which the server accepts client connections CFG=/opt/zookeeper/conf/zoo.cfg.dynamic CFG_BAK=/opt/zookeeper/conf/zoo.cfg.bak MY_ID_FILE=/tmp/zookeeper/myid HOSTNAME=$(hostname) while read -ra LINE; do PEERS=("${PEERS[@]}" $LINE) done # Don't add the first member as an observer if [ ${#PEERS[@]} -eq 1 ]; then # We need to write our index in this list of servers into MY_ID_FILE. # Note that this may not always coincide with the hostname id. echo 1 > "${MY_ID_FILE}" echo "server.1=${PEERS[0]}:2888:3888;2181" > "${CFG}" # TODO: zkServer-initialize is the safe way to handle changes to datadir # because simply starting will create a new datadir, BUT if the user changed # pod template they might end up with 2 datadirs and brief split brain. exit fi # Every subsequent member is added as an observer and promoted to a participant echo "" > "${CFG_BAK}" i=0 LEADER=$HOSTNAME for peer in "${PEERS[@]}"; do let i=i+1 if [[ "${peer}" == *"${HOSTNAME}"* ]]; then MY_ID=$i MY_NAME=${peer} echo $i > "${MY_ID_FILE}" echo "server.${i}=${peer}:2888:3888:observer;2181" >> "${CFG_BAK}" else if [[ $(echo srvr | /opt/nc "${peer}" 2181 | grep Mode) = "Mode: leader" ]]; then LEADER="${peer}" fi echo "server.${i}=${peer}:2888:3888:participant;2181" >> "${CFG_BAK}" fi done # zookeeper won't start without myid anyway. # This means our hostname wasn't in the peer list. if [ ! -f "${MY_ID_FILE}" ]; then exit 1 fi # Once the dynamic config file is written it shouldn't be modified, so the final # reconfigure needs to happen through the "reconfig" command. cp ${CFG_BAK} ${CFG} # TODO: zkServer-initialize is the safe way to handle changes to datadir # because simply starting will create a new datadir, BUT if the user changed # pod template they might end up with 2 datadirs and brief split brain. /opt/zookeeper/bin/zkServer.sh start # TODO: We shouldn't need to specify the address of the master as long as # there's quorum. According to the docs the new server is just not allowed to # vote, it's still allowed to propose config changes, and it knows the # existing members of the ensemble from *its* config. ADD_SERVER="server.$MY_ID=$MY_NAME:2888:3888:participant;0.0.0.0:2181" /opt/zookeeper/bin/zkCli.sh reconfig -s "${LEADER}":2181 -add "${ADD_SERVER}" # Prove that we've actually joined the running cluster ITERATION=0 until $(echo config | /opt/nc localhost 2181 | grep "${ADD_SERVER}" > /dev/null); do echo $ITERATION] waiting for updated config to sync back to localhost sleep 1 let ITERATION=ITERATION+1 if [ $ITERATION -eq 20 ]; then exit 1 fi done /opt/zookeeper/bin/zkServer.sh stop