The API provides two generic XML-RPC methods: one for applying changes to resources (apply()) and one for querying for resource information (query()).

apply([resource_record, …])

apply() accepts a list of resource records. It is guaranteed that either all changes will be applied or, when an error occurs, an XML-RPC fault will be returned and no changes will have been applied. On success the result will be a list of resource records indicating the new state of the resources.

Resource records are mappings that specify:

The type of the resource (see Resource Types). Mandatory.
The action to perform (set or delete). If __action__ is ommitted then the API will assume set as default.
any other key
defines a value specific to the __type__.
>>> server.apply([
...     {'__type__': 'virtualmachine', 'name': 'test00', 'cores': 1},
...     {'__type__': 'virtualmachine', '__action__': 'delete', 'name': 'test01'}])
[{'__type__': 'virtualmachine', 'name': 'test00', 'cores': 1, 'memory': 512, ...},

Resource records may be sparse: only keys that are given will be updated, others will be kept. You have to pass at least the keys marked as primary keys for a resource, though. Sparse updates allow you to perform changes without having to first retrieve and then save the whole record. It also avoids race conditions that could be introduced through a non-transactional read/update cycle. The result of apply() will always show the full resource records with all values as they were seen at the end of the transaction.

If you create new resources through the apply() method you need to provide enough data to make the record identifiable (all primary key fields) and additional fields that are marked as required.

Deletions may be processed immediately or delayed. For example, virtual machines will be deleted over a period of time according to our retention schedule. Those objects will show that they have been scheduled for deletion in their resource record.

Also, deleting resources that do not exist is not an error. Deleted or non-existing objects will not have a result in the response.

query(type, filter=None)

The query() method returns all resources of a given type that are visible to the current access key. Without any filters this includes the data of all child projects.

The list of valid types is given in the section Resource Types.

Filters are specified as a dictionary selecting type-specific keys from the result for exact matches:

>>> server.query('virtualmachine', {'name': 'test00'})
[{'__type__': 'virtualmachine', 'name': 'test00', 'cores': 1, 'memory': 512, ...}]


The log() method returns all log entries about any call to the other methods within the same project. This allows you to integrate our API logging into your own logging infrastructure by calling this method regularly. A serial counter assures that you will receive only new logs and not miss any logs.

A few rules:

  • Logs are kept 30 days.
  • Calls to the log() method are not logged themselves.
  • If a project is deleted, then all associated logs are deleted immediately, too.
  • Please do not poll the API in extremely short cycles. Let’s say once every few minutes is sufficient.
  • Logs only contain calls from the exact API key associated with the project. Calls with access keys from child projects are not shown in the parent log.
  • Logs are kept when the API key is reset.

How to ensure that you receive a gapless log:

  • For the first call you can omit the optional parameter serial which will return the latest 10 records.
  • For future calls you should remember the highest serial that you have seen in the past and pass it by calling log(serial).

Here’s an example:

>>> server.query('virtualmachine')
>>> server.log()
[{'args': '{"args": ["virtualmachine"], "kw": {}}',
  'exception': '',
  'ipaddress': '',
  'method': 'query',
  'resource_group': 'services',
  'result': '[]',
  'serial': 2,
  'timestamp': '2015-12-23 01:02:03'}]
>>> server.log(2)


The data for args and result in the log are JSON-encoded strings.