[Этот блог был синдицирован с http://nitschinger.at/]
мотивация
Поскольку я работаю в составе инженерной группы в Couchbase, мне нужно выполнить свой код для различных серверных развертываний. Мы используем множество операционных систем и версий программного обеспечения, как и наши клиенты. Чтобы надежно исправлять ошибки и создавать новые функции, очень важно запустить и запустить кластер, максимально похожий на эти развертывания. Я знаю, что могу использовать все эти комбинации на EC2, но стоимость этого будет очень высокой, и большую часть времени она будет излишней.
Мне нужно, чтобы такой кластер заработал за несколько минут и не тратил слишком много времени на его настройку. Я слышал о Vagrant и Puppet в прошлом, но никогда не удосужился использовать их на своей собственной коробке (хотя я всегда использую VirtualBox на MacOS для создания виртуальных машин вручную).
Этим утром я сел, чтобы поближе взглянуть на то, как эти инструменты могут помочь мне стать более продуктивным, и к моему огромному удивлению, я получил кластер Couchbase Server с 4 узлами, работающий менее чем за 30 минут (с поиском всех деталей конфигурации) , Так как это так просто, я хочу поделиться с вами.
Предпосылки
Прежде чем мы сможем подготовить наши узлы, вы должны убедиться, что у вас установлены Vagrant и VirtualBox. Если вы используете MacOS, как я, просто загрузите файлы .dmg для обоих, и все готово. Теперь создайте каталог где-нибудь для хранения файлов конфигурации — я назвал мои «бродяги».
In this directory, you need to create a Vagrantfile. Its like the Vagrants makefile and it will pick it up to learn how you want to have your nodes provisioned. Note that this doesn’t configure the software on top of the OS (like installing Couchbase), this is handled by puppet in a separate step. Here is the full config:
Vagrant.configure("2") do |config|
# Number of nodes to provision
numNodes = 4
# IP Address Base for private network
ipAddrPrefix = "192.168.56.10"
# Define Number of RAM for each node
config.vm.provider "virtualbox" do |v|
v.customize ["modifyvm", :id, "--memory", 1024]
end
# Provision the server itself with puppet
config.vm.provision :puppet
# Download the initial box from this url
config.vm.box_url = "http://files.vagrantup.com/precise64.box"
# Provision Config for each of the nodes
1.upto(numNodes) do |num|
nodeName = ("node" + num.to_s).to_sym
config.vm.define nodeName do |node|
node.vm.box = "precise64"
node.vm.network :private_network, ip: ipAddrPrefix + num.to_s
node.vm.provider "virtualbox" do |v|
v.name = "Couchbase Server Node " + num.to_s
end
end
end
end
This file is just ruby code that configures Vagrant. Let’s go through each directive and see what it does for us.
# Number of nodes to provision numNodes = 4 # IP Address Base for private network ipAddrPrefix = "192.168.56.10"
You can change these values, I just created them to fit my environment here. Depending on the amount of `numNodes` set, VMs will be created. I added a loop down below depending on this setting, so I don’t have to duplicate code a lot. The ip address prefix is used to easily determine the (static) IP address for the server. The numbers will be counted upwards incrementally, so you will end up with four servers accessible through `192.168.56.101` to `192.168.56.104`.
# Define Number of RAM for each node config.vm.provider "virtualbox" do |v| v.customize ["modifyvm", :id, "--memory", 1024] end
This config block is needed to increase the memory size of the VM. By default its less than that (I believe around 512MB), and I want to have 1 gig of RAM for each. Of course, feel free to tune that value or remove it completely.
# Provision the server itself with puppet config.vm.provision :puppet
Because we’ll be using puppet to provision the server software, we need to tell Vagrant to use it.
# Download the initial box from this url config.vm.box_url = "http://files.vagrantup.com/precise64.box"
Vagrant reuses predefined images so you don’t have to reinstall everything from scratch. Here we use a predefined Ubuntu 12.04 64bit box.
# Provision Config for each of the nodes
1.upto(numNodes) do |num|
nodeName = ("node" + num.to_s).to_sym
config.vm.define nodeName do |node|
node.vm.box = "precise64"
node.vm.network :private_network, ip: ipAddrPrefix + num.to_s
node.vm.provider "virtualbox" do |v|
v.name = "Couchbase Server Node " + num.to_s
end
end
end
This code block configures each virtual machine. Given the number of nodes we want to create, for each of them it assigns an IP address and gives it a descriptive name inside Virtualbox. If you want to add server-dependent settings, the «node» block is the right place for it. Otherwise it will pick the cluster wide settings defined in the «config» block.
Now if we would run `vagrant up` from the command line in this directory, we’d get four Ubuntu machines setup where we could SSH into, but nothing else would be installed. In order to make them do something, we want to install Couchbase Server. Puppet is a system automation software and very good at provisioning systems. Vagrant has amazing support for it, all we need to is create a `default.pp` file inside a `manifests` directory that looks like this:
exec { "couchbase-server-source":
command => "/usr/bin/wget http://packages.couchbase.com/releases/2.0.1/couchbase-server-enterprise_x86_64_2.0.1.deb",
cwd => "/home/vagrant/",
creates => "/home/vagrant/couchbase-server-enterprise_x86_64_2.0.1.deb",
before => Package['couchbase-server']
}
exec { "install-deps":
command => "/usr/bin/apt-get install libssl0.9.8",
before => Package['couchbase-server']
}
package { "couchbase-server":
provider => dpkg,
ensure => installed,
source => "/home/vagrant/couchbase-server-enterprise_x86_64_2.0.1.deb"
}
Let’s go over the internals once more.
exec { "couchbase-server-source":
command => "/usr/bin/wget http://packages.couchbase.com/releases/2.0.1/couchbase-server-enterprise_x86_64_2.0.1.deb",
cwd => "/home/vagrant/",
creates => "/home/vagrant/couchbase-server-enterprise_x86_64_2.0.1.deb",
before => Package['couchbase-server']
}
We also need to install `libssl0.9.8` on the server, this is the only dependency it has. We use the command line tool apt-get for this.
package { "couchbase-server":
provider => dpkg,
ensure => installed,
source => "/home/vagrant/couchbase-server-enterprise_x86_64_2.0.1.deb"
}
Finally, we can install the debian package from couchbase-server, because the file is in place and all dependencies are satisfied.
Of course, this puppet file is very simple and I’m you can do much more with it (and maybe even simplify it more) — but for my needs it is more than enough. If I want a different server version, I just need to change the puppet file and point it to the new debian package.
Now if we run vagrant up again, much more happens. Note that if you want to play with your puppet files, you can also use vagrant provision to apply the changes while the node is running.
If everything is okay, the output should look like this:
Bringing machine 'node1' up with 'virtualbox' provider... Bringing machine 'node2' up with 'virtualbox' provider... Bringing machine 'node3' up with 'virtualbox' provider... Bringing machine 'node4' up with 'virtualbox' provider... [node1] Clearing any previously set forwarded ports... [node1] Creating shared folders metadata... [node1] Clearing any previously set network interfaces... [node1] Preparing network interfaces based on configuration... [node1] Forwarding ports... [node1] -- 22 => 2222 (adapter 1) [node1] Running any VM customizations... [node1] Booting VM... [node1] Waiting for VM to boot. This can take a few minutes. [node1] VM booted and ready for use! [node1] Configuring and enabling network interfaces... [node1] Mounting shared folders... [node1] -- /vagrant [node1] -- /tmp/vagrant-puppet/manifests [node1] Running provisioner: puppet... Running Puppet with default.pp... stdin: is not a tty notice: /Stage[main]//Exec[install-deps]/returns: executed successfully notice: Finished catalog run in 0.77 seconds .... more for all the other nodes.
You can then point your browser to 192.168.56.10[1-4] and work with your Couchbase cluster. If you are done with it, you can use the vagrant halt command to shut it down cleanly. Very handy is also vagrant suspend, which will save the state of the nodes instead of shutting them down completely.
If you want to interact with one of the nodes instead of the whole cluster, you can always specify the node identifier. For example, if you want to start only the first node you can do it with the vagrant up node1 command.
To me, this is a very fast and clean way to provision server nodes. I just need to change a few lines in a file and get a new cluster without much hassle. Even more important, I can put those config files in version control and share them with other folks!