Setting up multi nodes live migration in Openstack Juno with devstack

Summary

  • Live migration overview

  • Setup Openstack with devstack: one controller node, two compute nodes
  • Configure migrations
  • Live migration test
  • Issue summary

Live migration Overview

Migration enables an administrator to move a virtual-machine instance from one compute host to another. This feature is useful when a compute host requires maintenance. Migration can also be useful to redistribute the load when many VM instances are running on a specific physical machine.

The migration types are:

  • Non-live migration (sometimes referred to simply as 'migration'). The instance is shut down for a period of time to be moved to another hypervisor. In this case, the instance recognizes that it was rebooted.

  • Live migration (or 'true live migration'). Almost no instance downtime. Useful when the instances must be kept running during the migration. The different types of live migration are:

    • Shared storage-based live migration. Both hypervisors have access to shared storage.

    • Block live migration. No shared storage is required. Incompatible with read-only devices such as CD-ROMs and Configuration Drive (config_drive).

    • Volume-backed live migration. Instances are backed by volumes rather than ephemeral disk, no shared storage is required, and migration is supported (currently only available for libvirt-based hypervisors

This page's live migration uses block live migration which does not need shared storage.

Setup Openstack with devstack: one controller node, two compute nodes

Prepare three servers with ubuntu installed. One server is worked as control node, the other two are compute nodes(Note: please named three servers with different hostnames.) and get updates and install git

sudo apt-get update

sudo apt-get install git

Configure controller node

  1. Git devstack Juno release:  git clone https://github.com/openstack-dev/devstack.git -b stable/juno
  2. Add localrc under devstack and add content to it as following

    HOST_IP=your configured IP address
    MULTI_HOST=
    FIXED_RANGE=10.0.0.0/
    FIXED_NETWORK_SIZE=
    ADMIN_PASSWORD=
    MYSQL_PASSWORD=
    RABBIT_PASSWORD=
    SERVICE_PASSWORD=
    SERVICE_TOKEN=
    SCREEN_LOGDIR=/home/stack/log/screen
    LOGFILE=/home/stack/log/stack.log
    LOGDAYS=
    PUBLIC_INTERFACE=eth0 #actual name of your network interface
    FLAT_INTERFACE=eth0
  3. Run devstack to install openstack: ./stack.sh

Configure compute node

  1. Git devstack Juno release:  git clone https://github.com/openstack-dev/devstack.git -b stable/juno
  2. Add localrc under devstack and add content as done in controller node. and also add following to it.

    SERVICE_HOST=your controller ip
    
    HOST_IP=host ip
    
    MULTI_HOST=
    
     FIXED_RANGE=10.0.0.0/
    
    FIXED_NETWORK_SIZE=
    
     ADMIN_PASSWORD=
    
    MYSQL_PASSWORD=
    
    RABBIT_PASSWORD=
    
    SERVICE_PASSWORD=
    
    SERVICE_TOKEN=
    
    SCREEN_LOGDIR=/home/stack/log/screen
    
    LOGFILE=/home/stack/log/stack.log
    
    LOGDAYS=
    
    #PUBLIC_INTERFACE=eth0
    
    FLAT_INTERFACE=eth0
    
    DATABASE_TYPE=mysql
    
    Q_HOST=$SERVICE_HOST
    
    MYSQL_HOST=$SERVICE_HOST
    
    RABBIT_HOST=$SERVICE_HOST
    
    GLANCE_HOSTPORT=$SERVICE_HOST:
    
    KEYSTONE_AUTH_HOST=$SERVICE_HOST
    
    KEYSTONE_SERVICE_HOST=$SERVICE_HOST
    
    ENABLED_SERVICES=n-cpu,n-net,n-api,c-sch,c-api,c-vol
  3. Add some variables which would use during install openstack
    Copy following info to a shell file:

    #!/bin/bash
    
    export OS_AUTH_URL=http://controller'ip:5000/v2.0
    export OS_TENANT_ID=69eaaebc44174418956d4dd104f2be76
    export OS_TENANT_NAME="admin" export OS_USERNAME="admin"
    export OS_PASSWORD=
    export OS_VOLUME_API_VERSION=

    Update the OS_AUTH_URL with your controller node ip. And use command:mysql -uroot -p123456 -e "select * from keystone.project;"  to get OS_TENANT_ID and update it, then run this shell

  4. Run devstack to install openstack: ./stack.sh

Configure migrations

  1. Host name configuration
    Add three server's host names to /etc/hosts to be sure that every node can ping successful to others

    server's IP   server's host name
  2. Configure /etc/sysconfig/iptables file to allow libvirt listen on TCP port 16509 and add a record accepting KVM communication on TCP port within the range from 49152 to 49261
    Note: ubuntu does not have sysconfig folder under etc, need create it manually.

    -A INPUT -p tcp -m multiport --ports  -m comment --comment "libvirt" -j ACCEPT
    
    -A INPUT -p tcp -m multiport --ports : -m comment --comment "migration" -j ACCEPT
  3. Configuration libvirt
    Enable libvirt listen flag at /etc/sysconfig/libvirtd file: add

    LIBVIRTD_ARGS="-listen"

    Configure /etc/libvirt/libvirtd.conf:

    listen_tls =
    listen_tcp =
    auth_tcp = "none"

    Configure /etc/init/libvirt-bin.conf, modify exec /usr/sbin/libvirtd  to exec /usr/sbin/libvirtd -l
    Configure /etc/libvirt/qemu.conf,  modify security_driver="none"

    Restart libvirt, after executing the command, ensure that libvirt is successfully restarted: sudostop libvirt-bin && sudo start libvirt-bin

    ps -ef |grep libvirt

  4. Nova configuration
    Modify:force_config_drive = None(value from always to None)

    Add:live_migration_flag=VIR_MIGRATE_UNDEFINE_SOURCE,VIR_MIGRATE_PEER2PEER,VIR_MIGRATE_LIVE  (under default)

  5. Restart service on each node
    screen -s stack -X quit
    ./rejoin-stack.sh

Live migration test

Before start live migration test, to be sure that the nova compute service are ready

      

  1. create instance and attach volume
    nova boot --flavor=flavor_id --image=image_id instance_name
    nova volume-attach vm_id volume_id auto

    +----------+--------------------------------------+
    | Property | Value |
    +----------+--------------------------------------+
    | device | /dev/vdb |
    | id | 3deb171f-8d61-4df9-8e6e-29d86575e78e |
    | serverId | bb12f3b0-c17e-49f7-b482-255258931eb4 |
    | volumeId | 3deb171f-8d61-4df9-8e6e-29d86575e78e |
    +----------+--------------------------------------+
  2. How to mount the volume which attached to the instance
    1> execute command to ssh to the instance, for example: ssh cirros@10.0.0.2
    2> execute command to list the partition tables for the installed devices: sudo fdisk -l
         see:Disk /dev/vdb......
    3> create a filesystem on the  device: sudo mkfs.ext4 /dev/vdb
    4> create a dictionary and mount the volume:
         sudo mkdir /data
         sudo mount /dev/vdb /data/
    5> then we can create folder or file on it
  3. nova show instance_id
  4. nova live-migration --block_migrate vm_id target_server_hostname
  5. check the host before and after migration

Issue summary

1. n-api could not start when install openstack with devstack
Solution: Restart controller node and rejoin the service, then run unstack and stack on compute node

2. Get error:libvirtError: operation failed: Failed to connect to remote libvirt URI qemu+tcp://computer247/system: Unable to resolve address 'computer247' service '16509': Name or service not known
Solution: check the iptables in configure migration step 2.

3. Live Migration failure: Invalid value '4-7,12-15' for 'cpuset.cpus': Invalid argument
Solution: Check the CPU info in your compute nodes, if there have different architecture of cpus, it would be fail to migrate

4. Compute node's nova-cpu service could not enable:


Solution: Check the libvirt settings in configure migration section, and restart libvirt-bin service, then run command in both compute node and controller node:
screen -s stack -X quit
./rejoin-stack.sh

5. Get error:


Solution: It's about nova configuration, check /etc/nova/nova.conf:
Modify:force_config_drive = None(value from always to None)      Add:live_migration_flag=VIR_MIGRATE_UNDEFINE_SOURCE,VIR_MIGRATE_PEER2PEER,VIR_MIGRATE_LIVE  (under default)
then rejoin stack in compute node

6. Getting error:screen-n-cpu.2015-06-02-184950.log:2015-06-04
00:16:45.797 ERROR nova.virt.libvirt.driver [-] [instance:
d1b56987-4691-4259-897f-ba4ce3e71260] Live Migration failure: Failed to
open file '/dev/disk/by-path/ip-9.115.246.45:3260-iscsi-iqn.1986-03.com.ibm:2145.v7k41.node1-lun-4': No such file or directory
Solution:  No solution currently by me, it is a bug and would be fixed

7. Gettting
error when rejoin stack:2015-06-09 14:02:33.154 TRACE
cinder.openstack.common.threadgroup ArgumentError: Could not parse
rfc1738 URL from string ''"


Solution:
modify the node sql_connection  in /etc/cinder/cinder.conf and
/etc/nova/nova.conf with( It has no value as default): take following as
example
/etc/cinder/cinder.conf:  sql_connection = mysql://root:123456@9.115.246.185/cinder?charset=utf8
/etc/nova/nova.conf:   sql_connection = mysql://root:123456@9.115.246.185/nova?charset=utf8

Setting up multi nodes live migration in Openstack Juno with devstack的更多相关文章

  1. Deploying Cloud Foundry on OpenStack Juno and XenServer (Part I)

    link http://rabbitstack.github.io/deploying-cloud-foundry-on-openstack-juno-and-xenserver-part-i/ Cl ...

  2. 使用openshit在ubuntu14.04下一键部署openstack(juno版本)

    一.基本介绍 本实验是在vmware workstation上虚拟机ubuntu14.04(64bit,desktop)上部署openstack(Juno版本).采用的工具是openshit.open ...

  3. openStack juno for ubuntu12-04

    <一,preinstall basic conf,pre Env> 1,pwgen(openssl rand -hex 10) some Open-Stack services add a ...

  4. OpenStack Juno 版本发布——支持Spark和NFV[转]

    作者:郑晨,OpenStack中国社区,转载请注明出处 美国时间2014年10月16日,OpenStack Juno版本正式发布,这是OpenStack开源云计算项目自2010年创立以来的第10个版本 ...

  5. 一键安装openstack juno 之controller node.

    原文名称: OpenStack Juno Scripted Installation on CentOS 7 Step I:  本机信息配置 CONTROLLER_IP=192.168.173.133 ...

  6. [openStack]使用Fuel安装OpenStack juno的fuel_master

    安装OpenStack是一件很复杂的事情,特别是在想目中,如果一个组件一个组件,一台一台的coding部署,估计太消耗时间,而且出错的概率很高,所以使用工具推送部署的效率就很高了,而且必须得可靠.mi ...

  7. vmware vms migration to openstack

    Converting a VMware Workstation virtual machine to KVM Leave a commentPosted by rbgeek on August 13, ...

  8. VMware migration to openstack kvm

  9. Centos7 install Openstack Juno (RDO) (转载)

    原文地址:http://www.hdume.com/centos-7-0%E5%AE%89%E8%A3%85openstack/ 1.安装系统,Centos7镜像采用CentOS-7.0-1406-x ...

随机推荐

  1. Gini 系数与熵的关系

    首先来看二者的基本定义: ⎧⎩⎨⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪H(X)=−∑k=1KpklnpkGini(X)=∑k=1Kpk(1−pk) 将 f(x)=−lnx 在 x=1 处进行一阶泰勒展开(忽略高阶无穷小 ...

  2. 学习OpenCV研究报告指出系列(二)源代码被编译并配有实例project

    下载并安装CMake3.0.1       要自己编译OpenCV2.4.9的源代码.首先.必须下载编译工具,使用的比較多的编译工具是CMake. 以下摘录一段关于CMake的介绍: CMake是一个 ...

  3. Maven项目中读取src/main/resources目录下的配置文件

    在Maven项目的开发中,当需要读取src/下的配置文件时,该怎么做? 我们假设Resources下有一个文件名为kafka.properties的配置文件(为什么用kafka.properties, ...

  4. PHP基本的文件和文件夹操作常用的汇总

    资源:http://www.ido321.com/835.html 一.基本文件的操作 文件的基本操作有:文件推断.文件夹推断.文件大小.读写性推断.存在性推断及文件时间等 1: <?php 2 ...

  5. mysql升级到5.7时间戳(timestamp)默认值报错

    原文:mysql升级到5.7时间戳报错 往数据库里创建新表的时候报错: [Err] 1067 - Invalid default value for 'updateTime' DROP TABLE I ...

  6. CSS(网页样式语言)基础

    所谓全栈,个体可以独立地完成系统的设计.开发.测试.部署以及运维.打通一个领域从无到有的全过程. 为什么会有 markdown 文本编辑显示工具呢,因为 html 太重了.markdown 是 htm ...

  7. s3c2410 cs8900a 网卡驱动程序

    /* CS8900a.h */ #define CONFIG_CERF_CS8900A 1 /* * cs8900a.c: A Crystal Semiconductor (Now Cirrus Lo ...

  8. matplotlib:path effects

    import matplotlib.pyplot as plt, matplotlib.patheffects as path_effects 1. normal fig = plt.figure(f ...

  9. 使用Ocelot做网关

    1首先创建一个json的配置文件,文件名随便取,我取Ocelot.json 这个配置文件有两种配置方式,第一种,手动填写 服务所在的ip和端口:第二种,用Consul进行服务发现 第一种如下: { & ...

  10. WPF的消息机制(一)- 让应用程序动起来

    原文:WPF的消息机制(一)- 让应用程序动起来 版权声明:本文为博主原创文章,未经博主允许不得转载. https://blog.csdn.net/powertoolsteam/article/det ...