To start the agents on a node, a configuration file must exist on the node at /etc/flocker/agent.yml.
The file must always include version and control-service items, and will need to include dataset objects similar to these:
"version": 1
"control-service":
"hostname": "${CONTROL_NODE}"
"port": 4524
# The dataset key below selects and configures a dataset backend (see below: aws/openstack/etc).
# All nodes will be configured to use only one backend
dataset:
backend: "aws"
region: "<your region; for example, us-west-1>"
zone: "<your availability zone; for example, us-west-1a>"
access_key_id: "<AWS API key identifier>"
secret_access_key: "<Matching AWS API key>"
The value of the hostname field should be a hostname or IP that is routable from all your node agents.
When configuring node agents, consider whether the location you choose for the Flocker control service will have multiple possible addresses, and ensure the hostname you provide is the correct one.
Warning
You should never choose 127.0.0.1 or localhost as the hostname, even if the control service is on same machine as the node agent, as this will keep the control service from correctly identifying the agent’s IP address.
It is also important to be aware that the flocker nodes will refuse to communicate with the flocker agent if there is a misconfiguration in the hostname. Please ensure that your hostname is configured correctly before proceeding, because any errors can result in failures.
Please note that the interface you choose will be the one that Flocker control traffic will be routed over. If you’re in environment where some interfaces have bandwidth costs and some are free (for example, AWS), ensure that you choose the private interface where bandwidth costs don’t apply.
${CONTROL_NODE} should be replaced with the address of the control node.
The optional port variable is the port on the control node to connect to.
This value must agree with the configuration for the control service telling it on what port to listen.
Omit the port from both configurations and the services will automatically agree.
The dataset item selects and configures a dataset backend.
All nodes must be configured to use the same dataset backend.
For a full list of available storage backends, see Supported Backends.
Warning
You can only choose a single backend at a time, and changing backends is not currently supported.
Each backend provides configuration details for you to setup the driver. When your driver is configured, you will need to return to complete the Flocker installation process.