ine.api.model.4.5.2.source-code.003-CommonConcepts.adoc Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of model Show documentation
Show all versions of model Show documentation
Model management tools for the oVirt Engine API.
== Common concepts
=== Types
The API uses the _type_ concept to describe the different kinds of objects
accepted and returned.
There are three relevant kinds of types:
Primitive types:: Describe simple kinds of objects, like <> or
<>.
Enumerated types:: Describe lists of valid values like <>
or <>.
Structured types:: Describe structured objects, with multiple attributes and
links, like <> or <>.
=== Identified types
Many of the types used by the API represent _identified_ objects, objects
that have an unique identifier and exist independently of other objects.
The types used to describe those objects extend the <>
type, which contains the following set of common attributes:
[cols="20,20,60"]
|===
|Attribute |Type |Description
|`id`
|<>
|Each object in the virtualization infrastructure contains an `id`, which
acts as an unique identifier.
|`href`
|<>
|The canonical location of the object as an absolute path.
|`name`
|<>
|A user-supplied human readable name for the object. The `name` name is
unique across all objects of the same type.
|`description`
|<>
|A free-form user-supplied human readable description of the object.
|===
IMPORTANT: Currently for most types of objects the `id` attribute is actually a
randomly generated https://en.wikipedia.org/wiki/Universally_unique_identifier[UUID],
but this is an implementation detail, and users should not rely on that, as
it may change in the future. Instead users should assume that these
identifiers are just strings.
=== Objects
Objects are the individual instances of the types supported by the API.
For example, the virtual machine with identifier `123` is an object of
the <> type.
=== Collections
A collection is a set of objects of the same type.
=== Representations
The state of objects needs to be represented when it is transferred
beetween the client and the server. The API supports XML and JSON as the
representation of the state of objects, both for input and output.
==== XML representation
The XML representation of an object consists of an XML element
corresponding to the type of the object, XML attributes for the `id` and
`href` attributes, and nested XML elements for the rest of the
attributes. For example, the XML representation for a virtual machine
appears as follows:
[source,xml]
----
myvm
My VM
1073741824
...
----
The XML representation of a collection of objects consists of an XML
element, named after the type of the objects, in plural. This contains
the representations of the objects of the collection. For example, the
XML respresentation for a collection of virtual machines appears as
follows:
[source,xml]
----
yourvm
Your VM
1073741824
...
myname
My description
2147483648
...
...
----
IMPORTANT: In the XML representation of objects the `id` and `href`
attributes are the only ones that are represented as XML attributes, the
rest are represented as nested XML elements.
==== JSON representation
The JSON representation of an object consists of a JSON document
containing a name/value pair for each attribute (including `id` and
`href`). For example, the JSON representation of a virtual machine
appears as follows:
....
{
"id": "123",
"href": "/ovirt-engine/api/vms/123",
"name": "myvm",
"description": "My VM",
"memory": 1073741824,
...
}
....
The JSON representation of a collection of objects consists of a JSON
document containg a name/value pair (named ater the type of the objects,
in singular) which in turn contains an array with the representations of
the objects of the collection. For example, the JSON respresentation for
a collection of virtual machines appears as follows:
....
{
"vm": [
{
"id": "123",
"href": "/ovirt-engine/api/vms/123",
"name": "myvm",
"description": "My VM",
"memory": 1073741824,
...
},
{
"id": "456",
"href": "/ovirt-engine/api/vms/456",
"name": "yourvm",
"description": "Your VM",
"memory": 2147483648,
...
},
]
}
....
=== Services
Services are the parts of the server responsible for retrieving, adding
updating, removing and executing actions on the objects supported by the
API.
There are two relevant kinds of services:
Services that manage a collection of objects:: These services are
reponsible for listing existing objects and adding new objects. For
example, the <> service is responsible for managing
the collection of virtual machines available in the system.
Services that manage a specific object:: These services are responsible
for retrieving, updating, deleting and executing actions in specific
objects. For example, the <> service is responsible for
managing a specific virtual machine.
Each service is accessible via a particular _path_ within the server.
For example, the service that manages the collection of virtual machines
available in the system is available in the via the path `/vms`, and the
service that manages the virtual machine `123` is available via the path
`/vms/123`.
All kinds of services have a set of _methods_ that represent the
operations that they can perform. The services that manage collections
of objects usually have the `list` and `add` methods. The services that
manage specific objects usually have the `get`, `update` and `remove`
methods. In addition, services may also have _action_ methods, that
represent less common operations. For example, the <>
service has a <> method that is used
to start a virtual machine.
For the more usual methods there is a direct mapping between the name of
the method and the name of the HTTP method:
[cols="50,50"]
|===
|Method name |HTTP method
|`add` |POST
|`get` |GET
|`list` |GET
|`update` |PUT
|`remove` |DELETE
|===
The path used in the HTTP request is the path of the service, with the
`/ovirt-engine/api` prefix.
For example, the request to `list` the virtual machines should be like
this, using the HTTP `GET` method and the path `/vms`:
....
GET /ovirt-engine/api/vms
....
For action methods the HTTP method is always `POST`, and the name of the
method is added as a suffix to the path. For example, the request to
start virtual machine `123` should look like this, using the HTTP `POST`
method and the path `/vms/123/start`:
....
POST /ovirt-engine/api/vms/123/start
....
Each method has a set of parameters.
Parameters are classified into two categories:
Main parameter:: The main parameter corresponds the object or collection
that is retrieved, added or updated. This only applies to the `add`,
`get`, `list` and `update` methods, and there will be exactly one such
main parameter per method.
Secondary parameters:: The rest of the parameters.
For example, the operation that adds a virtual machine (see
<>) has three parameters: `vm`, `clone`
and `clone_permissions`. The main parameter is `vm`, as it describes the
object that is added. The `clone` and `clone_permissions` parameters are
secondary parameters.
The main parameter, when used for input, must be included in the body of
the HTTP request. For example, when adding a virtual machine, the `vm`
parameter, of type <>, must be included in the request
body. So the complete request to add a virtual machine, including all
the HTTP details, must look like this:
....
POST /ovirt-engine/api/vms HTTP/1.1
Host: myengine.example.com
Authorization: Bearer fqbR1ftzh8wBCviLxJcYuV5oSDI=
Content-Type: application/xml
Accept: application/xml
myvm
My VM
Default
Blank
....
When used for output, the main parameters are included in the response
body. For example, when adding a virtual machine, the `vm` parameter
will be included in the response body. So the complete response body
will look like this:
....
HTTP/1.1 201 Created
Content-Type: application/xml
myvm
My VM
...
....
Secondary parameters are only allowed for input (except for action
methods, which are described later), and they must be included as query
parameters. For example, when adding a virtual machine with the `clone`
parameter set to `true`, the complete request must look like this:
....
POST /ovirt-engine/api/vms?clone=true HTTP/1.1
Host: myengine.example.com
Authorization: Bearer fqbR1ftzh8wBCviLxJcYuV5oSDI=
Content-Type: application/xml
Accept: application/xml
myvm
My VM
Default
Blank
....
Action methods only have secondary parameters. They can be used for
input and output, and they should be included in the request body,
wrapped with an `action` element. For example, the action method used to
start a virtual machine (see <>) has a
`vm` parameter to describe how the virtual machine should be started,
and a `use_cloud_init` parameter to specify if
https://cloudinit.readthedocs.io[cloud-init] should be used to configure
the guest operating system. So the complete request to start virtual
machine `123` using _cloud-init_ will look like this when using XML:
....
POST /ovirt-engine/api/vms/123/start HTTP/1.1
Host: myengine.example.com
Authorization: Bearer fqbR1ftzh8wBCviLxJcYuV5oSDI=
Content-Type: application/xml
Accept: application/xml
true
eth0
true
static
192.168.0.100
255.255.255.0
192.168.0.1
192.168.0.1
....
=== Searching
The `list` method of some services has a `search` parameter that can be
used to specify search criteria. When used, the server will only return
objects within the collection that satisfy those criteria. For example,
the following request will return only the virtual machine named `myvm`:
....
GET /ovirt-engine/api/vms?search=name%3Dmyvm
....
==== Maximum results parameter
Use the `max` parameter to limit the number of objects returned. For
example, the following request will only return one virtual machine,
regardless of how many are available in the system:
....
GET /ovirt-engine/api/vms?max=1
....
A search request without the `max` parameter will return all the
objects. Specifying the `max` parameter is recommended to reduce the
impact of requests in the overall performance of the system.
==== Case sensitivity
By default queries are not case sensitive. For example, the following
request will return the virtual machines named `myvm`, `MyVM` and `MYVM`:
....
GET /ovirt-engine/api/vms?search=name%3Dmyvm
....
The optional `case_sensitive` boolean parameter can be used to change this
behaviour. For example, to get exactly the virtual machine named `myhost`, and
not `MyHost` or `MYHOST`, send a request like this:
....
GET /ovirt-engine/api/vms?search=name%3D=myvm&case_sensitive=true
....
==== Search syntax
The `search` parameters use the same syntax as the {product-name} query
language:
....
(criteria) [sortby (element) asc|desc]
....
The `sortby` clause is optional and only needed when ordering results.
Example search queries:
[cols="20,20,60"]
|===
|Collection |Criteria |Result
|`hosts`
|`vms.status=up`
|Returns a list of all hosts running virtual machines that are `up`.
|`vms`
|`domain=example.com`
|Returns a list of all virtual machines running on the specified domain.
|`vms`
|`users.name=mary`
|Returns a list of all virtual machines belonging to users with the user
name `mary`.
|`events`
|`severity > normal sortby time`
|Returns a list of all events with severity higher than `normal` and
sorted by the the value of their `time` attribute.
|`events`
|`severity > normal sortby time desc`
|Returns a list of all events with severity higher than `normal` and
sorted by the the value of their `time` attribute in descending order.
|===
The value of the `search` parameter must be
https://en.wikipedia.org/wiki/Percent-encoding[URL-encoded] to translate
reserved characters, such as operators and spaces. For example, the
equal sign should be encoded as `%3D`:
....
GET /ovirt-engine/api/vms?search=name%3Dmyvm
....
==== Wildcards
The asterisk can be used as part of a value, to indicate that any string
matches, including the emtpy string. For example, the following request
will return all the virtual machines with names beginning with `myvm`,
such as `myvm`, `myvm2`, `myvma` or `myvm-webserver`:
....
GET /ovirt-engine/api/vms?search=name%3Dmyvm*
....
==== Pagination
Some {product-name} environments contain large collections of objects.
Retrieving all of them with one request isn't practical, and hurts
performace. To allow retrieving them page by page the `search` parameter
supports an optional `page` clause. This, combined with the `max`
parameter, is the basis for paging. For example, to get the first page
of virtual machines, with a page size of 10 virtual machines, send
request like this:
....
GET /ovirt-engine/api/vms?search=page%201&max=10
....
NOTE: The search parameter is URL-encoded, the actual value of the
`search` parameter, before encoding, is `page 1`, so this is actually
requesting the first page.
Increase the `page` value to retrieve the next page:
....
GET /ovirt-engine/api/vms?search=page%202&max=10
....
The `page` clause can be used in conjunction with other clauses inside
the `search` parameter. For example, the following request will return
the second page of virtual machines, but sorting by name:
....
GET /ovirt-engine/api/vms?search=sortby%20name%20page%202&max=10
....
[IMPORTANT]
====
The API is stateless; it is not possible to retain a state between
different requests since all requests are independent from each other.
As a result, if a status change occurs between your requests, then the
page results may be inconsistent.
For example, if you request a specific page from a list of virtual
machines, and virtual machines are created or removed before you request
the next page, then your results may be missing some of them, or contain
duplicates.
====
[id="documents/003_common_concepts/follow"]
=== Following links
The API returns references to related objects as _links_. For example, when
a virtual machine is retrieved it contains links to its disk attachments
and network interface cards:
[source,xml]
----
...
...
----
The complete description of those _linked_ objects can be retrieved by sending
separate requests:
....
GET /ovirt-engine/api/vms/123/diskattachments
GET /ovirt-engine/api/vms/123/nics
....
However, in some situations it is more convenient for the application using the API to
retrieve the linked information in the same request. This is useful, for example,
when the additional network round trips introduce an unacceptable overhead, or when
the multiple requests complicate the code of the application in an unacceptable
way. For those use cases the API provides a `follow` parameter that allows the
application to retrieve the linked information using only one request.
The value of the `follow` parameter is a list of strings, separated by commas.
Each of those strings is the _path_ of the linked object. For example, to retrieve
the disk attachments and the NICs in the example above the request should be like
this:
....
GET /ovirt-engine/api/vms/123?follow=disk_attachments,nics
....
That will return an response like this:
[source,xml]
----
...
true
true
virtio_scsi
false
false
false
...
eth0
virtio
true
00:1a:4a:16:01:00
true
...
...
----
The path to the linked object can be a single word, as in the previous example,
or it can be a sequence of words, separated by dots, to request nested
data. For example, the previous example used `disk_attachments` in order to
retrieve the complete description of the disk attachments, but each
disk attachment contains a link to the disk, which wasn't _followed_.
In order to also follow the links to the disks, the following request
can be used:
....
GET /ovirt-engine/api/vms/123?follow=disk_attachments.disk
....
That will result in the following response:
[source,xml]
----
true
true
virtio_scsi
false
false
false
mydisk
My disk
0
raw
true
ok
image
0
...
...
...
----
The path can be made as deep as needed. For example, to also get the statistics
of the disks:
....
GET /ovirt-engine/api/vms/123?follow=disk_attachments.disk.statistics
....
Multiple path elements and multiple paths can be combined. For example, to get the
disk attachments and the network interface cards, both with their statistics:
....
GET /ovirt-engine/api/vms/123?follow=disk_attachments.disk.statistics,nics.statistics
....
IMPORTANT: Almost all the operations that retrieve objects support the `follow`
parameter, but make sure to explicitly check the reference documentation, as
some operations may not support it, or may provide advice on how to use
it to get the best performance.
IMPORTANT: Using the `follow` parameter moves the overhead from the
client side to the server side. When you request additional data, the
server must fetch and merge it with the basic data. That consumes CPU
and memory in the server side, and will in most cases require additional
database queries. That may adversely affect the performance of the server,
especially in large scale environments. Make sure to test your application
in a realistic environment, and use the `follow` parameter only when
justified.
=== Permissions
Many of the services that manage a single object provide a reference to
a `permissions` service that manages the permissions assigned to that
object. Each permission contains links to the user or group, the role
and the object. For example, the permissions assigned to a specific
virtual machine can be retrieved sending a request like this:
....
GET /ovirt-engine/api/vms/123/permissions
....
The response body will look like this:
[source,xml]
----
...
----
A permission is added to an object sending a `POST` request with a
permission representation to this service. Each new permission requires
a role and a user.
=== Handling errors
Some errors require further explanation beyond a standard HTTP status
code. For example, the API reports an unsuccessful object state update
or action with a `fault` in the response body. The fault contains the
`reason` and `detail` attributes. For example, when the server receives
a request to create a virtual machine without the mandatory `name`
attribute it will respond with the following HTTP response line:
....
HTTP/1.1 400 Bad Request
....
And the following response body:
[source,xml]
----
Incomplete parameters
Vm [name] required for add
----