作者:张华 发表于:2012-3-30
版权声明:可以任意转载,转载时请务必以超链接形式标明文章原始出处和作者信息及本版权声明
( http://blog.csdn.net/quqi99 )
1网络桥接配置
首先,fedora16对网卡使用了新的命名方式,第一块网卡不叫eth0,我的T420上叫em1。它为什么做,有兴趣可参考文献:
Fedora15的新的网卡命名方式http://www.20ju.com/content/V168885.htm
总之,我不喜欢这种叫法,现在将网卡名称从em1改回eth0,具体做法请参考文献,(修改fedora16的网络接口名,使其变回eth0http://www.bitbi.biz/fedora-2/%E4%BF%AE%E6%94%B9fedora-16-%E7%9A%84%E7%BD%91%E7%BB%9C%E6%8E%A5%E5%8F%A3%E5%90%8D%EF%BC%8C%E4%BD%BF%E5%85%B6%E5%8F%98%E5%9B%9E-eth0-2685)
最后配置桥接(参考文献:http://blog.bodhizazen.net/linux/fedora-15-virt-manager-bridged-networking/):
sudoyuminstallbridge-utils
sudovi/etc/sysconfig/network-scripts/ifcfg-eth0
DEVICE=eth0
TYPE=Ethernet
#changethehardwareaddresstomatchthehardwareaddressyourNICuses
HWADDR=00:21:CC:60:D6:1A
ONBOOT=yes
#BOOTPROTO=dhcp
#MODE=Managed
BRIDGE=br100
NM_CONTROLLED=no
sudovi/etc/sysconfig/network-scripts/ifcfg-br100
DEVICE=br100
TYPE=Bridge
BOOTPROTO=dhcp
#BOOTPROTO=static
#IPADDR=192.168.99.100
#NETWORK=192.168.99.0
#NETMASK=255.255.255.0
#GATEWAY=192.168.99.1
#DNS1=202.106.195.30
ONBOOT=yes
NM_CONTROLLED=no
最后重启网络服务,sudoservicenetworkrestart,需要注意的是,fedora默认使用NetworkManager来管理网卡,而briage不支持NetworkManager,所以需要关闭NetworkManager服务(sudoserviceNetworkManagerstop&sudoserviceNetworkManager disable)。不然会报下列错(tail-f/var/log/message):
http://forums.fedoraforum.org/showthread.php?t=262356
Bringingupinterfaceeth0:Error:Connectionactivationfailed:DevicenotmanagedbyNetworkManagerorunavailable
[FAILED]
RTNETLINKanswers:Fileexists
不过,我没有使用上述方法,因为你关闭NetworkManager服务的话,你的无线网卡也不能用NetworkManager管理了,所以我在上述的配置中加了“NM_CONTROLLED=no”实现了同样的目的。
我机器的其他条件是:
1)你得确保安装了libvirtd(yuminstalllibvirtd&servicelibvirtdstart),可用virsh-cqemu:///systemlist命令测试
2)安装gcc,这个在安装pythonmodules时会用到
3)mysql的root用户应该有密码,可用echo"showdatabases"|mysql-uroot-ppassword
4)在命令行中先切换到root用户,然后再启动eclipse
2通过git下载源码
1) 安装java与eclipse,略
2)为eclipse安装pydev插件(help->installnewsoftwae...,http://pydev.org/updates),然后再"window->preferences":菜单中的Pydev->Interpreter–Python选项卡中点击AutoConfig按钮配置python
3) 为eclipse安装egit插件(http://download.eclipse.org/egit/updates)
4)为eclipse安装django插件:
installdjangopluginforpython,firstdownloadhttp://www.djangoproject.com/download/1.3.1/tarball/
sudo
python
setup.py
installexport
PATH=/usr/lib/python2.7/site-packages/django/bin:$PATH
django-admin.py
–
version
django-admin.py
startproject
myFirstDjango
cd
myFirstDjango
sudo
chmod
777
*
./manage.py
runserver
http://127.0.0.1:8000/
5)在eclipse中通过git插件clone下列组件的master分支(cloneagitrepository),注意,有的组件会下载失败,只需要将https协议换成git协议即可。我的代码下载在/bak/openstack目录下。
#computeservice
NOVA_REPO=https://github.com/openstack/nova.git
#storageservice
SWIFT_REPO=https://github.com/openstack/swift.git
#imagecatalogservice
GLANCE_REPO=https://github.com/openstack/glance.git
#unifiedauthsystem(managesaccounts/tokens)
KEYSTONE_REPO=https://github.com/openstack/keystone.git
#awebsockets/html5orflashpoweredVNCconsoleforvminstances
NOVNC_REPO=https://github.com/cloudbuilders/noVNC.git
#djangopoweredwebcontrolpanelforopenstack
HORIZON_REPO=https://github.com/openstack/horizon.git
#pythonclientlibrarytonovathathorizon(andothers)use
NOVACLIENT_REPO=https://github.com/openstack/python-novaclient.git
#pythonkeystoneclientlibrarytonovathathorizonuses
KEYSTONECLIENT_REPO=https://github.com/openstack/python-keystoneclient
#quantumservice
QUANTUM_REPO=https://github.com/openstack/quantum
#quantumclient
QUANTUM=https://github.com/openstack/python-quantumclient
#Tempesttestsuite
https://github.com/openstack/tempest.git
#melangeservice
https://github.com/openstack/melange.git
#pythonmelangeclientlibrary
https://github.com/openstack/python-melangeclient.git
#swiftclient
https://github.com/chmouel/python-swiftclient.git
3在eclipse中导入工程
1)在"GitRepositories"视图上上步clone下来的每一个工程的"WorkingDirector"目录上右键点击选"importprojects"导入工程,再选“importasgeneralproject",直接见下图吧。
2)依次做完上步之后,然后可以进入java视图,在每个工程上点右键选"Pydev->setasPydevproject",即使是horizon工程也不例外,设置的是pydevproject,而不是django工程
3安装一些依赖包
install python-pip,nowpipisareplacementfor`easy_install`
sudo yumsearchpython-pip
sudo yuminstallpython-pip.noarch
sudo yuminfopython-pip.noarch
sudo yuminstallgcc;
sudo yuminstallpython-devel
sudo yuminstallpython-greenlet
sudo yuminstallMySQL-python
sudo yuminstalleuca2ools
sudo yuminstalldnsmasq
sudo yuminstallscsi-target-utils
install MQ
rabbitmqconfiguration
sudoyuminstallrabbitmq-server
sudoservicerabbitmq-serverstart
sudochkconfigrabbitmq-serveron
changetherabbitpasswordsincethedefaultis"guest"
sudorabbitmqctlchange_passwordguestpassword
sudorabbitmqctlstatus
install mysql
sudoyuminstallmysql-serverphp-mysql
sudoservicemysqldstart
sudochkconfigmysqldon
mysql-uroot-ppasswordSETPASSWORDFOR'root'@'localhost'=PASSWORD('password');
sudomysql-uroot-ppassword-h127.0.0.1-e"GRANTALLPRIVILEGESON*.*TO'root'@'%'identifiedby'password';"
sudocp/usr/share/mysql/my-small.cnf/etc/my.cnf
sudosed-i's/127.0.0.1/0.0.0.0/g'/etc/my.cnf
install libvirt
sudo yuminstalllibvirt
sudo servicelibvirtdrestart
6) sudoyuminstallpython-sqlite2python-lxmlpython-greenlet-develpython-ldap
4 安装依赖的Pythonmodules
依次检查,如果有的工程下的目录有tools/pip-requires,则可以运行下列命令安装依赖的pythonmodules,例:
cd/bak/openstack/python-novaclient/
pip-python install-rtools/pip-requires
pip-python install-rtools/test-requires
并对除了你要在eclipse中运行的工程(keystone,glance,nova,horion)除外的其他工程(novaclient,keystoneclient)依次运行命令:
python setup.py develop
注意,如果在对nova运行时报“ImportError:cannotimportnamenormalize”,normalize是/usr/lib64/python2.7/local.py里的一个方法,原因是nova工程的目录下有一个locale的目录将local的名称空间给盖了(http://groups.google.com/group/django-users/browse_thread/thread/e8bb9089d9e5be60)
5最小化配置openstack
只需要启动keystone,glance-reg,glance-api,nova-all,dashboard四个组件就可以启动openstack了,其他的服务不是必须的。其中,由于dashboard是一个django工程,我们不需要安装apache,用django自带的http服务器(pythonmanage.pyrunserver)即可启动它。
5.1配置keystone
1)配置:
vi/bak/openstack/keystone/etc/keystone.conf
connection=mysql://root:password@localhost/keystone
driver=keystone.token.backends.sql.Token
driver=keystone.contrib.ec2.backends.sql.Ec2
2)创建数据库:
mysql-uroot-ppassword-e'dropdatabasekeystone'
mysql-uroot-ppassword-e'createdatabasekeystone'
mysql-uroot-ppassword-e"grantallon*.*toroot@'%'identifiedby'password'"
cd/bak/openstack/keystone&&./bin/keystone-managedb_sync
3)启动
若是在命令行中如下方法启动
./bin/keystone-all--config-file /bak/openstack/keystone/etc/keystone.conf --log-config/bak/openstack/keystone/etc/logging.conf -d –debug
若是在eclipse中,如下图,一定要将Arguments的workingdirctory中的${workspace_loc:keystone/bin}改为{workspace_loc:keystone},如果你不想这样改的话,那就你就得在Programarguments那段填参数--config-file/bak/openstack/keystone/etc/keystone.conf --log-config/bak/openstack/keystone/etc/logging.conf -d –debug
初始化数据
SERVICE_TOKEN=ADMIN
ADMIN_PASSWORD=password
SERVICE_ENDPOINT=http://localhost:35357/v2.0
exportSERVICE_TOKEN=$SERVICE_TOKEN
exportSERVICE_ENDPOINT=$SERVICE_ENDPOINT
functionget_id(){
echo`$@|grepid|awk'{print$4}'`
}
ADMIN_TENANT=`get_idkeystonetenant-create--name=admin`
ADMIN_USER=`get_idkeystoneuser-create--name=admin--pass="$ADMIN_PASSWORD"--email=admin@example.com`
ADMIN_ROLE=`get_idkeystonerole-create--name=admin`
KEYSTONEADMIN_ROLE=`get_idkeystonerole-create--name=KeystoneAdmin`
KEYSTONESERVICE_ROLE=`get_idkeystonerole-create--name=KeystoneServiceAdmin`
keystoneuser-role-add--user$ADMIN_USER--role$ADMIN_ROLE--tenant_id$ADMIN_TENANT
keystoneuser-role-add--user$ADMIN_USER--role$KEYSTONEADMIN_ROLE--tenant_id$ADMIN_TENANT
keystoneuser-role-add--user$ADMIN_USER--role$KEYSTONESERVICE_ROLE--tenant_id$ADMIN_TENANT
echo$?
5)验证
keystoneservice-list
keystoneuser-list
keystonerole-list
架构http://keystone.openstack.org/architecture.html
Service,使用keystone的内部服务
Identity,基于角色的验证与授权
Token,user/tenant的credentials已经验证之后生成的随机数
Catalog,注册的后端,像sql,kvs,ldap
Policy,提供rule-basedauthorizationengine
nosetests-s-vtest_backend_sql.py
keystoneservice-create--nameCOMPUTE_ID--typecompute--description'OpenStackComputeservice'
keystoneendpoint-create--region=RegionOne--service_id=COMPUTE_ID--publicurl='http://localhost:$(compute_port)s/v1.1/$(tenant_id)s'--internalurl='http://localhost:$(compute_port)s/v1.1/$(tenant_id)s'–adminurl='http://localhost:$(compute_port)s/v1.1/$(tenant_id)s'
keystoneendpoint-create--region=RegionOne--service_id=IDENTITY_ID--publicurl='http://localhost:$(public_port)s/v2.0'--internalurl='http://localhost:$(public_port)s/v2.0'--adminurl='http://localhost:$(admin_port)s/v2.0'
keystoneendpoint-create--region=RegionOne--service_id=IMAGE_ID--publicurl='http://localhost:9292/v1'--internalurl='http://localhost:9292/v1'--adminurl='http://localhost:9292/v1'
执行上述语句要keystone.conf文件中有:
driver= keystone.catalog.backends.sql.Catalog
5.2配置glance
1)配置
vi/bak/openstack/glance/etc/glance-registry.conf
sql_connection=mysql://root:password@localhost/glance
vi/bak/openstack/glance/etc/glance-api-paste.conf
vi/bak/openstack/glance/etc/glance-registry-paste.conf
#admin_tenant_name=%SERVICE_TENANT_NAME%
#admin_user=%SERVICE_USER%
#admin_password=%SERVICE_PASSWORD%
admin_token=ADMIN
vi/bak/openstack/glance/etc/glance-api.conf
vi/bak/openstack/glance/etc/glance-registry.conf
[paste_deploy]
flavor=keystone
debug=True
#log_file=/var/log/glance/api.log
2)数据库:
mysql-uroot-ppassword-e'DROPDATABASEIFEXISTSglance;'
mysql-uroot-ppassword-e'CREATEDATABASEglance;'
mysql-uroot-ppassword-e"grantallon*.*toroot@'%'identifiedby'password'"
cd/bak/openstack/glance&&./bin/glance-managedb_sync
3)启动
./bin/glance-registry--config-file=/bak/openstack/glance/etc/glance-registry.conf
filesystem_store_datadir=/var/lib/glance/images
./bin/glance-api–config-file=/bak/openstack/glance/etc/glance-api.conf
如果是在eclipse中启动,注意和上面keystone的一样,将基准路径从${workspace_loc:glance/bin}改为${workspace_loc:glance},这样改了之后,就直接从${workspace_loc:glance/}/etc/glance目录读配置文件,而不是从/etc/glance读
4) 上传镜像
cd /bak/openstack/glance
export FILES=/bak/openstack/glance
mkdir -p $FILES/images
exportOS_AUTH_USER=admin
export OS_AUTH_KEY=password
exportOS_AUTH_TENANT=admin
exportOS_AUTH_URL=http://localhost:35357/v2.0
exportOS_AUTH_STRATEGY=keystone
sudo wget -chttp://images.ansolabs.com/tty.tgz -O $FILES/images/tty.tgz
tar-zxf $FILES/images/tty.tgz -C $FILES/images
TOKEN=`curl -s -d"{\"auth\":{\"passwordCredentials\":{\"username\": \"$OS_AUTH_USER\", \"password\":\"$OS_AUTH_KEY\"}, \"tenantName\":\"$OS_AUTH_TENANT\"}}" -H "Content-type:application/json" http://localhost:5000/v2.0/tokens | python -c"import sys; import json; tok = json.loads(sys.stdin.read());print tok['access']['token']['id'];"`
RVAL=`./bin/glanceadd -A $TOKEN name="cirros-kernel" is_public=truecontainer_format=aki disk_format=aki <$FILES/images/aki-tty/image`
KERNEL_ID=`echo $RVAL | cut -d":"-f2 | tr -d " "`
RVAL=`./bin/glance add-A $TOKEN name="cirros-ramdisk" is_public=truecontainer_format=ari disk_format=ari<$FILES/images/ari-tty/image`
RAMDISK_ID=`echo $RVAL | cut-d":" -f2 | tr -d " "`
./bin/glance add -A$TOKEN name="cirros" is_public=true container_format=amidisk_format=ami kernel_id=$KERNEL_ID ramdisk_id=$RAMDISK_ID<$FILES/images/ami-tty/image
5) 验证
./bin/glance-f index
5.4nova配置
sudoyuminstallscsi-target-utils
sudoservicetgtdrestart(如果启动不了,可换用于fedora17的tgtdrpm包)
sudosystemctlstatustgtd.service
sudovgsnova-volumes
sudolosetup-f--show/bak/kvmimages/nova.img
sudovgcreatenova-volumes/dev/loop0
sudoservicetgtdrestart
sudosystemctlstatustgtd.service
sudosystemctlstatusnetwork.service
实例路径:/bak/openstack/nova/instances
查看实例:virshlist--all
查看iscsisession:iscsiadm--modenode
删除iscsisession:sudoiscsiadm--modenode--logout&&sudoiscsiadm--modenode--opdelete
1)配置文件
vi/etc/nova/nova.conf
--verbose
--logdir=/var/log/nova
--state_path=/var/lib/nova
--lock_path=/var/lib/nova/tmp
--instances_path=/var/lib/nova/instances
--debug=True
--scheduler_driver=nova.scheduler.simple.SimpleScheduler
--dhcpbridge_flagfile=/etc/nova/nova.conf
--fixed_range=10.0.0.0/24
--s3_host=127.0.0.1
--network_manager=nova.network.manager.FlatDHCPManager
--volume_group=nova-volumes
--volume_name_template=volume-%08x
--iscsi_helper=tgtadm
--osapi_compute_extension=nova.api.openstack.compute.contrib.standard_extensions
--my_ip=localhost
--public_interface=br100
--vlan_interface=eth0
--sql_connection=mysql://root:password@localhost/nova
--libvirt_type=kvm
--instance_name_template=instance-%08x
--novncproxy_base_url=http://127.0.0.1:6080/vnc_auto.html
--xvpvncproxy_base_url=http://127.0.0.1:6081/console
--vncserver_listen=127.0.0.1
--vncserver_proxyclient_address=127.0.0.1
--api_paste_config=/etc/nova/api-paste.ini
--image_service=nova.image.glance.GlanceImageService
--ec2_dmz_host=127.0.0.1
--rabbit_host=localhost
--rabbit_password=password
--glance_api_servers=127.0.0.1:9292
--force_dhcp_release
--connection_type=libvirt
--firewall_driver=nova.virt.libvirt.firewall.IptablesFirewallDriver
--flat_network_bridge=br100
--flat_interface=eth0
--auth_strategy=keystone
cp/bak/openstack/nova/etc/nova/api-paste.ini/etc/nova/
vi/bak/openstack/nova/etc/nova/api-paste.ini
#admin_tenant_name=%SERVICE_TENANT_NAME%
#admin_user=%SERVICE_USER%
#admin_password=%SERVICE_PASSWORD%
admin_token=ADMIN
2)数据库
mysql-uroot-ppassword-e'DROPDATABASEIFEXISTSnova;'
mysql-uroot-ppassword-e'CREATEDATABASEnova;'
cd/bak/openstack/nova&&sudo./bin/nova-managedbsync
3)fixthepydevproblemwith"eventlet.monkey_patch()"
cd/bak/openstack/nova/bin
perl-pi-e's/eventlet\.monkey_patch\(\)/eventlet\.monkey_patch\(all=False,socket=True,select=True\)/g'*
4)千万不要将glance安装到系统中,但要在eclipse中设置glance及keystone做nove的工程依赖,如果已安装,可按下列方法清除。
pip-pythonuninstallglance
rm-rf/usr/bin/glance*
rm-rf/usr/lib/python2.7/site-packages/glance.egg-link
5)启动
vi/bak/openstack/nova/bin/nova-all
#forbinaryin['nova-compute','nova-network','nova-scheduler','nova-cert']:
forbinaryin['nova-compute','nova-network','nova-scheduler']:
启动:/bak/openstack/nova/bin/nova-all
5)验证
#forkeystoneclient
exportSERVICE_ENDPOINT=http://localhost:35357/v2.0
exportSERVICE_TOKEN=ADMIN
#fornovaclient
exportNOVA_AUTH_STRATEGY=keystone
exportNOVA_USERNAME=admin
exportNOVA_PASSWORD=password
exportNOVA_PROJECT_ID=admin
exportNOVA_URL=http://localhost:35357/v2.0
exportNOVA_VERSION=1.1
#for glanceclient
export OS_AUTH_USER=admin
exportOS_AUTH_KEY=password
export OS_AUTH_TENANT=admin
exportOS_AUTH_URL=http://localhost:35357/v2.0
exportOS_AUTH_STRATEGY=keystone
nova list
nova flavor-list
nova image-list
nova –debugboot--flavorm1.tiny--image5213ac41-a588-419f-9b50-c86217433df9quqitestserver
nova-manageservicelist
若网络服务启动失败,可能要:rm-rf/var/lib/nova/tmp/nova-iptables.lock
如果上述命令不能执行,有可能是因为它无法找到policy.json文件,需要将eclipse的基准主路设为${workspace_loc:nova/etc/nova}
5.3配置dashboard
mv /bak/openstack/horizon/openstack_dashboard/local /bak/openstack/horizon/openstack_dashboard/local/local_settings.py
vi/bak/openstack/horizon/openstack_dashboard/local/local_settings.py
DATABASES= {
'default':{
'ENGINE':'django.db.backends.mysql',
'NAME':'dash',
'USER':'root',
'PASSWORD':'password',
'HOST':'localhost',
'default-character-set':'utf8',
},
}
vi/bak/openstack/horizon/manage.py
#!/usr/bin/envpython
fromdjango.core.management import execute_manager
importos, sys
ROOT_PATH= os.path.dirname(os.path.abspath(__file__))
ifROOT_PATH not in sys.path:
sys.path.insert(0,ROOT_PATH)
importopenstack_dashboard.settings
if__name__ == "__main__":
execute_manager(openstack_dashboard.settings)
1)创建dashboard数据库
sudo
mysql
-uroot
-ppassword
-e
'create
database
dash';
sudomysql-uroot-ppassword-e"grantallon*.*toroot@'%'identifiedby'password'"
setPYTHONPATH=/bak/openstack/glance:/bak/openstack/keystone
cd/bak/openstack/horizon/
pythonmanage.pysyncdb
配置它工程依赖glance,keystone
将它自身的顶级目录openstack_dashboard加入到PYTHONPATH
启动,cd/bak/openstack/horizon
pythonmanage.py runserver ,用它启动horion时,就可以不需要启动apache
2012-04.05添加:
上述配置是可以正常启动实例的,但是实例没有IP,那是因为忘了建立虚拟网络,命令如下:
nova-manage network createpublic10.10.10.0/24 1 256
其他测试用到的命令如下:删除虚拟网络:nova-managenetworkdelete 10.10.10.0/24启动实例:nova boot --image 5213ac41-a588-419f-9b50-c86217433df9--flavor1 --nic net-id=99b1d65e-34ae-4658-8387-ce9724 myserver8
删除实例:novadelete myserver8
给实例加硬盘:nova volume-create--display_namevolhua11(1代表1G)
删除硬盘:novavolume-attach myserver8 2 /dev/vdb
另外,测试了一下分布式部署,基本上和上面是一样的,注意以下几点就行了:
1)之前单机部署时控制节点的/etc/nova/nova.conf中使用的localhost可以不用动。因为本来mysql, activemq, glance都是安装在控制节点上。
2)但是控制节点 glance/etc/default_catalog.tempates中的IP需要从localhost变为实际IP
3)在计算节点上的/etc/nova/nova.conf中的IP相应改对,如mysql, activemq, glance的IP
4)计算节点仅仅只需要启动一个nova-compute即可工作。但是nova-compute运行时需要通过nova-client, keystone-client, glance-client去调用相应的服务。但是glance为client弄一个工程,所以在计算节点也得像上文中的方法安装nova-client, keystone-client, glance三个依赖工程
另外,如果在eclipse中无法调试nova工程的话,那是因为nova中使用了一个叫做eventlet的python协程库(大家知道的nginx是java里的一个非阻塞I/O库),所谓协程,是一个比线程更小的单位,协程运行在用户空间,所以操作系统无法感知它,所以它需要程序员自己去调度协程,这样用于执行协作式多任务效率比只用线程的更高。eventlet中有一个叫monkey patch的东东,它用来在运行时对己有的代码进行修改,达到hot patch的目的。eventlet使用该技巧用来替换一些python的标准的库,比如说socket。就是因为这个东东把一些标准库给替换成它自己的库然后与pydev插件冲突,所以造成无法在eclipse中调试程序。比如说现在想调试bin/nova-network,那样可以修改bin/nova-network脚本,将evenlet.monkye_patch()替换成:
eventlet.monkey_patch(all=False,socket=True,select=True)