mesos入门（三）—— HA模式

搭建高可用的mesos时对原来的项目还是有很大规模的修改的，同时还修改了一些以前遗留的bug

简介

机器环境

[all]
192.168.50.4
192.168.50.5
192.168.50.6
192.168.50.7

[master]
192.168.50.4
192.168.50.5
192.168.50.6

[slave]
192.168.50.4
192.168.50.5
192.168.50.6
192.168.50.7

zookeepr+mesos-master+marathon均部署在master中，mesos-slave+docker均部署在slave中

具体搭建可以看https://github.com/ncuwaln/mesos-learn

接下来说一下搭建中的问题

问题

• zookeepr myid

问题描述：zookeepr需要动态创建每一个zookeepr的myid文件以及其中的内容，接下来就是如何用ansible从zoo.cfg中抽取出当前的主机的id

# 创建myid file
- name: Make id file
  file: path={{remote_dir}}/zookeeper/data/myid state=touch

# 获得本机IP，获得的IP用于从zoo.cfg中匹配id
# ps1: grep eth1是我的网卡，你的编号可能不同
# ps2: cut命令 -d选项是分隔符，-f是分割后的第几个区间的字符串
- name: get ip
  shell: ip addr|grep eth1|grep inet|awk '{print $2}'| cut -d / -f 1
  register: local_ip

# 根据ip匹配id
# ps: 重点还是cut命令的巧用
- name: get id
  shell: "grep {{local_ip['stdout']}} {{remote_dir}}/zookeeper/conf/zoo.cfg|cut -d \\= -f 1|cut -d \\. -f 2"
  register: myid

# debug用可注释
- name: echo
  debug: msg={{myid}}

# 将id写入
- name: write id
  lineinfile: path={{remote_dir}}/zookeeper/data/myid line={{myid['stdout']}}

• Mesos-master: Shutdown failed on fd=xx: Transport endpoint is not connected [107]

问题描述: Mesos-master: Shutdown failed on fd=xx: Transport endpoint is not connected [107]

启用mesos的advertise_ip选项

引用：https://stackoverflow.com/questions/33148588/mesos-master-shutdown-failed-on-fd-25-transport-endpoint-is-not-connected-107

• marathon只有leader节点的服务才可访问

问题描述： 只有leader的marathon服务的8080端口才可访问，其它机器的8080端口均503

启动marathon时添加hostname选项，非leader节点的服务才可以重定向到leader节点

mesos与marathon启动脚本

为了启动以及停止mesos与marathon方便，我编写了它们两个的启动脚本，仓库zookeeper的启动脚本
mesos.sh

#!/usr/bin/env bash

MESOSBINDIR="$( cd "$( dirname "$0"  )" && pwd  )"


MASTER_WORK_DIR="/data/mesos/master"
MASTER_LOG_DIR="/data/mesos/master/log"
SLAVE_WORK_DIR="/data/mesos/slave"
SLAVE_LOG_DIR="/data/mesos/slave/log"

USAGE=" hostname and advertise_ip quorum zk is reuired \n
--hostname <hostname> \n
--advertise_ip <advertise_ip> \n
--quorum \n
--zk"

hostname=""
advertise_ip=""
quorum=""
zk=""
master=""


case "$1" in
  start_master )
    while [[ -n "$2" ]]; do
      case "$2" in
        --hostname ) hostname=$3; shift 2;;
        --advertise_ip ) advertise_ip=$3; shift 2;;
        --quorum ) quorum=$3; shift 2;;
        --zk ) zk=$3; shift 2;;
        * ) break;;
      esac
    done
    if [ "$advertise_ip" = "" -o "$hostname" = "" -o "$quorum" = "" -o "$zk" = "" ]; then
      echo "error options"
      exit -1
    fi
    echo -n "Staring mesos-master ..."
    nohup "${MESOSBINDIR}/mesos-master" "--hostname=$hostname" "--advertise_ip=$advertise_ip" \
    "--quorum=$quorum" "--work_dir=$MASTER_WORK_DIR" "--zk=$zk" "--log_dir=$MASTER_LOG_DIR" &
    echo "started"
    ;;
  stop_master )
    pid=`ps -ef|grep mesos-master|grep -v "grep"|awk '{print $2}'`
    if [ "$pid" = "" ]; then
      echo "No mesos master server started"
      exit 0
    fi
    kill -9 $pid
    echo "Mesos master server stoped"
    ;;
  restart_master )
    shift
    "$0" stop_master ${@}
    sleep 5
    "$0" start_master ${@}
    ;;
  start_slave )
    while [[ -n "$2" ]]; do
      case "$2" in
        --hostname ) hostname=$3; shift 2;;
        --advertise_ip ) advertise_ip=$3; shift 2;;
        --master ) master=$3; shift 2;;
        * ) break;;
      esac
    done
    if [ "$advertise_ip" = "" -o "$hostname" = "" -o "$master" = "" ]; then
      echo -n "error options"
      exit -1
    fi
    echo "Starting mesos slave server ..."
    nohup "${MESOSBINDIR}/mesos-agent" "--hostname=$hostname" "--advertise_ip=$advertise_ip" \
    "--work_dir=$SLAVE_WORK_DIR" "--master=$master" "--log_dir=$SLAVE_WORK_DIR" &
    echo "started"
    ;;
  stop_slave )
    pid=`ps -ef|grep mesos-agent|grep -v "grep"|awk '{print $2}'`
    if [ "$pid" = "" ]; then
      echo "No mesos slave server started"
      exit 0
    fi
    kill -9 $pid
    echo "Mesos slave server stoped"
    ;;
  restart_slave )
    shift
    "$0" stop_slave ${@}
    sleep 5
    "$0" start_slave ${@}
    ;;
  * )
    echo -e $USAGE
    ;;
esac

marathon.sh

#!/usr/bin/env bash

MARATHONBINDIR="$( cd "$( dirname "$0"  )" && pwd  )"

USAGE=" master and zk is reuired \n
--master \n
--zk"

master=""
zk=""
libmesos_path=""
hostname=""


case "$1" in
  start )
    while [[ -n "$2" ]]; do
      case "$2" in
        --master ) master=$3; shift 2;;
        --zk ) zk=$3; shift 2;;
        --libmesos_path ) libmesos_path=$3; shift 2;;
        --hostname ) hostname=$3; shift 2;;
        * ) break;;
      esac
    done
    if [ "$master" = "" -o "$zk" = "" -o "$hostname" = ""]; then
      echo "error options"
      exit -1
    fi
    echo -n "Staring mesos-master ..."
    if [ ["$libmesos_path" = ""] ]; then
      nohup "${MARATHONBINDIR}/marathon" "--master" "$master" "--zk" "$zk" "--hostname" "$hostname"&
    else
      export MESOS_NATIVE_JAVA_LIBRARY=${libmesos_path}
      nohup "${MARATHONBINDIR}/marathon" "--master" "$master" "--zk" "$zk" "--hostname" "$hostname"&
    fi
    echo "started"
    ;;
  stop )
    pid=`ps -ef|grep marathon|grep -v "grep"|awk '{print $2}'`
    if [ "$pid" = "" ]; then
      echo "No marathon server started"
      exit 0
    fi
    kill -9 $pid
    echo "Mesos master server stoped"
    ;;
  restart )
    shift
    "$0" stop ${@}
    sleep 5
    "$0" start ${@}
    ;;
esac

脚本还有一点小bug，即启动前没判断是否已存在进程，下次commit时应该会一并更改吧，接下来的文章就是在HA模式的环境下的应用部署操作了