one-pf-manage

Simple platform management scripts using opennebul cli

objective

Describe a bunch of VM with an expressive and flexible DRY json structure, and do scripted operations (create/delete/check) based on that description

install

Python 3 script, uses standard modules only : json, logging, os, re, subprocess, xml.etree.ElementTree

Requires OpenNebula CLI tools (oneuser, onevm and onetemplate) for which you can read the official installation instructions.

Tested with OpenNebula virtual sandbox (using version 5.4)

how does it work

An exemple is provided in the docs folder.

The principle is the following :

• the platform_name defines the prefix for all VM name. It CANNOT be empty, or every accessible VM in OpenNebula would be considered part of the platform.
• the hosts element lists all desired VMs ; each VM name will be prefixed with platform_name, followed by a dash -, followed by the host name

Each host has a "desired" configuration, based on a succession of overrides, with the following precedence :

• initial configuration is read from defaults
• if the host references a class, the configuration is updated with the overrides defined in given class
• the configuration is finally updated with the eventual overrides found in the given host definition

With the following JSON content :

{
    "format_version":"3",
    "platform_name":"project-version",
    "defaults":{
        "one_template": null,
        "cpu_percent": 0.1,
        "vcpu_count": 1,
        "mem_mb": 128,
        "disks": [{
        "image": "ttylinux",
        "size_mb": 256
        }],
        "networks":["cloud"]
    },
    "classes":{
        "more_mem":{ "mem_mb": 512 },
        "with_template":{ "one_template": "ttylinux" },
        "recursive":{ "class": "more_mem" }
    },
    "hosts":{
        "srv1":{
            "class": "more_mem",
            "mem_mb": 384,
            "networks":[]
        },
        "srv2":{ "class": "with_template" },
        "srv3":{ "vcpu_count": 2 },
        "srv4":{},
        "srv5":{
            "mem_mb": 384,
            "class": "recursive"
        }
    }
}

This yields the following target configurations :

name: project-version-srv1
    cpu: 0.1
    vcpu: 1
    mem_mb: 384
    one_template: None
    networks: 0
    disks: 1
        image ttylinux of size 256 Mbytes
name: project-version-srv2
    cpu: 0.1
    vcpu: 1
    mem_mb: 128
    one_template: ttylinux
    networks: 1
        cloud
    disks: 1
        image ttylinux of size 256 Mbytes
name: project-version-srv3
    cpu: 0.1
    vcpu: 2
    mem_mb: 128
    one_template: None
    networks: 1
        cloud
    disks: 1
        image ttylinux of size 256 Mbytes
name: project-version-srv4
    cpu: 0.1
    vcpu: 1
    mem_mb: 128
    one_template: None
    networks: 1
        cloud
    disks: 1
        image ttylinux of size 256 Mbytes
name: project-version-srv5
    cpu: 0.1
    vcpu: 1
    mem_mb: 384
    one_template: None
    networks: 1
        cloud
    disks: 1
        image ttylinux of size 256 Mbytes

As you can see :

• the vm name uses the platform name as prefix
• project-version-srv5 classes are applied depth-first
• project-version-srv4 has no overrides at all and equals default
• project-version-srv3 had only a host override for vcpu
• project-version-srv2 had only a class override for one_template
• project-version-srv1 had both a class and a host override both modifying mem_mb and its final value respects precedence, and a host override for the network.

pre-requisites

As this tool uses the standard OpenNebula CLI, your environment variables ONE_XMLRPC must be configured appropriately, for use by the CLI tools.

You must pre-log into your OpenNebula cluster using the CLI tools (eventually defining ONE_AUTH and using oneuser) before running the provided script

usage

The -h option displays available command-line options.

You can start your json template, using docs/example.json and the above explanation as an example.

In case you have trouble or wonder what gets parsed out of your description, you can use the parse-only option.

Then you run ./opm.py status yourfile.json

For the example configuration, this yields :

$ ./opm.py status docs/example.json
project-version-srv1: missing
project-version-srv2: missing
project-version-srv3: missing
project-version-srv4: missing
project-version-srv5: missing
project-version-srv6: missing
project-version-srv7: missing

This means that the VM are not created. You can create them :

$ ./opm.py create-missing docs/example.json
project-version-srv1: created ID 43
project-version-srv2: created ID 44
project-version-srv3: created ID 45
project-version-srv4: created ID 46
project-version-srv5: created ID 47
project-version-srv6: created ID 48
project-version-srv7: created ID 49

Note : each VM is created on hold to allow for possible pxe boot menu. As a consequence, each VM must be released before it starts running. See your OpenNebula documentation for hold/release operations.

If you then add another host srv8 into the file, and run status :

$ ./opm.py status docs/example.json
project-version-srv8: missing
project-version-srv1: present ID 43
project-version-srv2: present ID 44
project-version-srv3: present ID 45
project-version-srv4: present ID 46
project-version-srv5: present ID 47
project-version-srv6: present ID 48
project-version-srv7: present ID 49

You can create the missing VM with the same creation command :

$ ./opm.py docs/example.json create-missing
project-version-srv8: created ID 50

If you remove a devinition from your file, and run status again :

$ ./opm.py status docs/example.json
project-version-srv1: present ID 43
project-version-srv2: present ID 44
project-version-srv3: present ID 45
project-version-srv4: present ID 46
project-version-srv5: present ID 47
project-version-srv7: present ID 49
project-version-srv8: present ID 50
project-version-srv6: unreferenced ID 48

You can remove the unreferenced VM using delete-unreferenced

$ ./opm.py delete-unreferenced docs/example.json
project-version-srv6: destroyed ID 48

If you update a definition in the json file, and run synchronize

$ ./opm.py synchronize docs/example.json
project-version-srv6: ID 64, changing cpu_percent from 0.1 to 0.2, changing vcpu_count from 1 to 2
project-version-srv7: ID 65, changing mem_mb from 96 to 64

Note: due to the risk of loss in case of bad behaviour, disk comparison and synchronization is not yet implemented and likely never will

Note: network is not yet implemented

Finally, you can remove all (existing) platform vm using delete-all

$ ./opm.py delete-all docs/example.json
project-version-srv1: destroyed ID 43
project-version-srv2: destroyed ID 44
project-version-srv3: destroyed ID 45
project-version-srv4: destroyed ID 46
project-version-srv5: destroyed ID 47
project-version-srv7: destroyed ID 49
project-version-srv8: destroyed ID 50

And voilà.