9.5 KiB

Raw Blame History

Sahara Default Template CLI

The sahara-templates application is a simple CLI for managing default templates in Sahara. This document gives an overview of default templates and explains how to use the CLI.

Default Templates Overview

The goal of the default template facility in Sahara is to make cluster launching quick and easy by providing users with a stable set of pre-generated node group and cluster templates for each of the Sahara provisioning plugins.

Template sets are defined in .json files grouped into directories. The CLI reads these template sets and writes directly to the Sahara database.

Default templates may only be created, modified, or deleted via the CLI --operations through the python-saharaclient or REST API are restricted.

JSON Files

Cluster and node group templates are defined in .json files.

A very simple cluster template JSON file might look like this:

{
    "plugin_name": "vanilla",
    "hadoop_version": "2.6.0",
    "node_groups": [
        {
            "name": "master",
            "count": 1,
            "node_group_template_id": "{master}"
        },
        {
            "name": "worker",
            "count": 3,
            "node_group_template_id": "{worker}"
        }
    ],
    "name": "cluster-template"
}

The values of the node_group_template_id fields are the names of node group templates in set braces. In this example, master and worker are the names of node group templates defined in .json files in the same directory. When the CLI processes the directory, it will create the node group templates first and then substitute the appropriate id values for the name references when it creates the cluster template.

Configuration Files and Value Substitutions

The CLI supports value substitution for a limited set of fields. For cluster templates, the following fields may use substitution:

default_image_id
neutron_management_network

For node group templates, the following fields may use substitution:

image_id
flavor_id
floating_ip_pool

Substitution is indicated for one of these fields in a .json file when the value is the name of the field in set braces. Here is an example of a node group template file that uses substitution for flavor_id:

{
    "plugin_name": "vanilla",
    "hadoop_version": "2.6.0",
    "node_processes": [
        "namenode",
        "resourcemanager",
        "oozie",
        "historyserver"
    ],
    "name": "master",
    "flavor_id": "{flavor_id}",
    "floating_ip_pool": "{floating_ip_pool}"
}

The values for flavor_id and floating_ip_pool in this template will come from a configuration file.

If a configuration value is found for the substitution, the value will be replaced. If a configuration value is not found, the field will be omitted from the template. (In this example, flavor_id is a required field of node group templates and the template will fail validation if there is no substitution value specifed. However, floating_ip_pool is not required and so the template will still pass validation if it is omitted).

The CLI will look for configuration sections with names based on the plugin_name, hadoop_version, and name fields in the template. It will look for sections in the following order:

[<name>]

May contain fields only for the type of the named template

If templates are named in an unambiguous way, the template name alone can be a used as the name of the config section. This produces shorter names and aids readability when there is a one-to-one mapping between template names and config sections.
[<plugin_name>_<hadoop_version>_<name>]

May contain fields only for the type of the named template

This form unambiguously applies to a specific template for a specific plugin.
[<plugin_name>_<hadoop_version>]

May contain node group or cluster template fields
[<plugin_name>]

May contain node group or cluster template fields
[DEFAULT]

May contain node group or cluster template fields

If we have the following configuration file in our example the CLI will find the value of flavor_id for the master template in the first configuration section and the value for floating_ip_pool in the third section:

[vanilla_2.6.0_master]
# This is named for the plugin, version, and template.
# It may contain only node group template fields.
flavor_id = 5
image_id = b7883f8a-9a7f-42cc-89a2-d3c8b1cc7b28

[vanilla_2.6.0]
# This is named for the plugin and version.
# It may contain fields for both node group and cluster templates.
flavor_id = 4
neutron_mangement_network = 9973da0b-68eb-497d-bd48-d85aca37f088

[vanilla]
# This is named for the plugin.
# It may contain fields for both node group and cluster templates.
flavor_id = 3
default_image_id = 89de8d21-9743-4d20-873e-7677973416dd
floating_ip_pool = my_pool

[DEFAULT]
# This is the normal default section.
# It may contain fields for both node group and cluster templates.
flavor_id = 2

Sample Configuration File

A sample configuration file is provided in sahara/plugins/default_templates/template.conf. This file sets the flavor_id for most of the node group templates supplied with Sahara to 2 which indicates the m1.small flavor in a default OpenStack deployment.

The master node templates for the CDH plugin have the flavor_id set to 4 which indicates the m1.large flavor, since these nodes require more resources.

This configuration file may be used with the CLI as is, or it may be copied and modified. Note that multiple configuration files may be passed to the CLI by repeating the --config-file option.

Other Special Configuration Parameters

The only configuration parameter that is strictly required is the connection parameter in the database section. Without this value the CLI will not be able to connect to the Sahara database.

By default, the CLI will use the value of the plugins parameter in the [DEFAULT] section on update to filter the templates that will be created or updated. This parameter in Sahara defaults to the set of fully supported plugins. To restrict the set of plugins for the update operation set this parameter or use the --plugin-name option.

Directory Structure

The structure of the directory holding .json files for the CLI is very flexible. The CLI will begin processing at the designated starting directory and recurse through subdirectories.

At each directory level, the CLI will look for .json files to define a set of default templates. Cluster templates may reference node group templates in the same set by name. Templates at different levels in the directory structure are not in the same set.

Plugin name and version are determined from the values in the .json files, not by the file names or the directory structure.

Recursion may be turned off with the "-n" option (see below).

The default starting directory is sahara/plugins/default_templates

Example CLI Commands

For update, delete, node-group-template-delete, and cluster-template-delete operations, the tenant must always be specified. For node-group-template-delete-id and cluster-template-delete-id tenant is not required. All useful information about activity by the CLI is logged

Create/update all of the default templates bundled with Sahara. Use the standard Sahara configuration file in /etc/sahara/sahara.conf to specify the plugin list and the database connection string and another configuration file to supply the flavor_id values:

$ sahara-templates --config-file /etc/sahara/sahara.conf --config-file myconfig update -t $TENANT_ID

Create/update default templates from the directory mypath:

$ sahara-templates --config-file myconfig update -t $TENANT_ID -d mypath

Create/update default templates from the directory mypath but do not descend into subirectories:

$ sahara-templates --config-file myconfig update -t $TENANT_ID -d mypath -n

Create/update default templates bundled with Sahara for just the vanilla plugin:

$ sahara-templates --config-file myconfig update -t $TENANT_ID -p vanilla

Create/update default templates bundled with Sahara for just version 2.6.0 of the vanilla plugin:

$ sahara-templates --config-file myconfig update -t $TENANT_ID -p vanilla -pv 2.6.0

Create/update default templates bundled with Sahara for just version 2.6.0 of the vanilla plugin and version 2.0.6 of the hdp plugin:

$ sahara-templates --config-file myconfig update -t $TENANT_ID -p vanilla -pv vanilla.2.6.0 -p hdp -pv hdp.2.0.6

Delete default templates for the vanilla plugin:

$ sahara-templates --config-file myconfig delete -t $TENANT_ID -p vanilla

Delete default templates for version 2.6.0 of the vanilla plugin:

$ sahara-templates --config-file myconfig delete -t $TENANT_ID -p vanilla -pv 2.6.0

Delete a specific node group template by ID:

$ sahara-templates --config-file myconfig node-group-template-delete-id --id ID

Delete a specific cluster template by ID:

$ sahara-templates --config-file myconfig cluster-template-delete-id --id ID

Delete a specific node group template by name:

$ sahara-templates --config-file myconfig node-group-template-delete --name NAME -t $TENANT_ID

Delete a specific cluster template by name:

$ sahara-templates --config-file myconfig cluster-template-delete-id --name NAME -t $TENANT_ID

9.5 KiB Raw Blame History