cookbook 'etcd-v2', '= 1.0.5', :supermarket
etcd-v2 (5) Versions 1.0.5 Follow1
Install and configure etcd
cookbook 'etcd-v2', '= 1.0.5'
knife supermarket install etcd-v2
knife supermarket download etcd-v2
This cookbook manages etcd version 2. It provides resources for installation from both binaries and source, and for management of one or more etcd service instnaces.
What even is all of this for?!
This cookbook aims to provide a feature complete utility to install and configure one or more instances of etcd for both production and development/testing proposes.
etcd_service resource allows multiple etcd processes to run on the same system to:
* Test client libraries with actual clusters. Not just the one node you
brew installd and started in another terminal.
* Test discovery configurations and failovers before doing it live.
* Figure out how to actually add/remove nodes to/from your cluster safely before you do this.
etcd_source resources ensure that their respecitve installations are vendored so as to allow mutliple versions/builds of etcd to coexist on a single system.
etcd_service resources each map to an installation, allowing you to test arbitrary compatability scenarios both within a cluster and with client libraries.
etcd-v2::aws Create a simple cluster using the
- etcd-v2::default Install base dependencies. Must be included before using resources.
- etcd-v2::node Create a standalone node
- etcd-v2::testing Create a cluster of three nodes on the same host for testing
Install etcd from a compiled release, by default from coreos/etcd on GitHub.
etcd_binary 'default' will install etcd in
/opt and link to
etcd_binary 'name' do version 'v2.0.10' # Default set in node['etcd_v2']['version'], currently 'v2.0.10' platform 'linux-amd64' # Default set in node['etcd_v2']['platform'] path '/opt' # Install root. Unpacked into <path>/etcd-<version>-<platform> srv_bin 'etcd' # Name of etcd binary in package ctl_bin 'etcdctl' # Name of etcd control binary in package bin_path '/usr/local/bin' # Path to link to binaries from. Set to nil or false to disable linking from :github # Package source. Currently only :github is supported repository '/coreos/etcd' # GitHub repo to fetch release from. Default node['etcd_v2']['bin_repository'] end
Fetch and build etcd from a git repository. This resource does not install golang. The golang cookbook should do the needful for you.
etcd_source 'name' do version 'v2.0.10' # Default set in node['etcd_v2']['version'], currently 'v2.0.10' path '/opt' # Install root. Unpacked into <path>/etcd-<version>-<platform>-<name> srv_bin 'etcd' # Name of etcd binary ctl_bin 'etcdctl' # Name of etcd control binary bin_path '/usr/local/bin' # Path to link to binaries from. Set to nil or false to disable linking repository 'email@example.com:coreos/etcd.git' # Git repo to fetch source from. Default node['etcd_v2']['source_repository'] end
Configure and run an installation (
etcd_source) as a service. The only required attributes in the following are
name_node (name attribute) and
instance. Other values have sane defaults for running a single node cluster.
While most configuration parameters are exposed directly, several abstractions are provided to capture some of the more confusing or repetitive parts of the etcd v2 configuration spec:
peer_host attributes are used to simplify the composition of various
*-url arguments. Arrays passed to these attributes will result in geometric compositions, including the
protocol arrtibute in the respective argument:
-advertise-client-urls\ http://client_host:client_port,\ http://client_host:client_port, http://client_host:client_port,\ ...
Static peers are added using the
peer(name, protocol, host, client_port, peer_port) method. The node's
-initial-cluster argument will be composed from a merge of
peer_port parameters as well as the nodes own 'peer_host:peer_port' set.
discovery attribute enables different configuration arguments specific to the respecive clustering method.
:dns are features of etcd. The
:aws discovery method is implemented by this cookbook. It uses the EC2 tags API to find peers for cluster bootstrapping.
etcd_service 'node_name' do node_name 'node0' # Name attribute. etcd node name ## Required: Installed version of etcd to use. Can be a string/hash reference or an instance ## of the `etcd_binary` or `etcd_source` resource (e.g. inherits Chef::Resource::Etcd). instance 'etcd_binary[default]' ## Service operator. User/group will be created if missing. Default 'etcd' user 'etcd' group 'etcd' ## Passed to unterlying `service` resource service_action [:start, :enable] ## Node configuration/tuning client_host 'localhost' # One or more advertised client hosts (String, Array). Default `node['ipaddress']` client_listen '0.0.0.0' # One or more client bind addresses (String, Array). Default `0.0.0.0` client_port 2379 # One or more client listen ports (Integer, Array). Default 2379 peer_host 'localhost' # One or more advertised peer hosts. Default `node['ipaddress']` peer_listen '0.0.0.0' # One or more peer bind addresses. Default `0.0.0.0` peer_port 2380 # One or more peer listen ports. Default 2380 ## See https://github.com/coreos/etcd/blob/master/Documentation/configuration.md#member-flags data_dir '/var/data/etcd' snapshot_count 10_000 max_snapshots 5 max_wals 5 heartbeat_interval 100 election_timeout 1000 cors 'allowed.domain.com' # CORS origins allowed by client API (String, Array). proxy :off # One of :on, :readonly, :off. See https://github.com/coreos/etcd/blob/master/Documentation/proxy.md ## SSL protocol :http # Transport protocol. One of :http, :https. Default :http ## See https://github.com/coreos/etcd/blob/master/Documentation/configuration.md#security-flags cert_file 'client-cert.pem' key_file 'client-key.pem' client_cert_auth true trusted_ca_file 'client-ca.pem' peer_cert_file 'peer-cert.pem' peer_key_file 'peer-key.pem' peer_client_cert_auth true peer_trusted_ca_file 'peer-ca.pem' ## Cluster configuration See https://github.com/coreos/etcd/blob/master/Documentation/configuration.md#clustering-flags for specifics discovery :static # One of :static, :etcd, :dns, :aws. quorum 3 # Resource will wait for a quorum to be available before configuring and starting etcd. Currently only used by `:aws` discovery state :new # initial-cluster-state: One of :new, :existing. Default :new token 'etcd-cluster' # initial-cluster-token: Default 'etcd-cluster' ## Define a peer node for :static configuration. peer 'node1', :http, 'localhost', 2381, 2382 discovery_service 'https://discovery.etcd.io/blahblahblah' # An etcd discovery node discovery_proxy 'proxy.domain.com' # HTTP(S) Proxy to etcd discovery servuce discovery_domain 'domain.com' # Domain inwhich to query DNS SRV record etcd-server[-ssl]._tcp.domain.com discovery_fallback :exit # One of :exit, :proxy. See https://github.com/coreos/etcd/blob/master/Documentation/proxy.md ## AWS Discovery parameters aws_tags :service => 'foo', :cluster => 'production' # Tags used to discover peers aws_host_attribute :private_dns_name # AWS-SDK Instance key to be used as hostnames See [The Docs](http://docs.aws.amazon.com/sdkforruby/api/Aws/EC2/Instance.html) ## AWS key attributes are provided for compatability with the Opscode AWS cookbook; ## however, **you should be using IAM instance profiles ಠ_ಠ** aws_access_key '' aws_secret_access_key '' end
Using the :aws discovery method
:aws discovery method uses AWS tags and etcd APIs to bootstrap a new cluster, join an exizting cluster, or fail back to a proxy node if a cluster of the desired size exists.
:awsdiscovery method requires the
awscookbook. You must add it to your downstream dependencies and include the
aws::defaultrecipe before defining resources that use the
:awsdiscovery method! The
aws::ec2_hintsrecipe may be necessary to coerce ohai into populating
node_namewill be set to
node['ec2']['instance_id']. The same transport
peer_portmust be used across the cluster.
- Due to the serial nature of Chef, multiple
etcd_serviceresources should not be defined in the same run_list for the same cluster when using the
:awsdiscovery method. One resource will block the Chef run until the desired quorum of peers is discovered. Note that the
:awsmethod uses the same underlying configuration as the
:staticmethod. For simple testing scenarios, they should be functionally equivalent.
- The instance will need the following authorizations. They should be configured in an IAM role and applied with an instance profile:
The included Vagrant file uses the
vagrant-aws plugins. If you would like to provision EC2 nodes, install both plugins and create a .vagrant/secret.yaml file. Run
vagrant secret-init and edit the resulting file:
# Account-specific AWS configurations access_key_id: 'xxx' secret_access_key: 'xxx' keypair_name: 'xxx' subnet_id: 'subnet-xxx' security_groups: - 'sg-xxx' # IAM Profile ARN. Associated role must have the actions, listed above, allowed. iam_instance_profile_arn: 'arn:aws:iam::xxx:instance-profile/xxx'
1.0.5 passed this metric
1.0.5 passed this metric