layout | title | published | sorting |
---|---|---|---|
default |
Augments |
true |
70 |
Augments files can be used to define variables and classes for use by all CFEngine components before any parsing or evaluation happen. Augments are fundamentally JSON data files, so you should view and edit them with a JSON-aware editor if possible. This is a convenient way to override defaults defined in the default policy, the Masterfiles Policy Framework (MPF), without modifying the shipped policy files.
As an example, you can add your own policy file to inputs and bundle name to the
bundle sequence, without editing promises.cf
, by adding the Augments file below
(/var/cfengine/masterfiles/def.json
):
{
"inputs": ["services/my_policy_file.cf"],
"vars":
{
"control_common_bundlesequence_end": ["my_bundle_name"]
}
}
In this case, the contents of the policy file, /var/cfengine/masterfiles/services/my_policy_file.cf
, could look something like this:
bundle agent my_bundle_name
{
files:
"/tmp/hello"
create => "true",
content => "cfengine";
}
You can ensure the file is deleted and use the info log output to confirm that the policy is actually being run:
$ cf-agent -Kf update.cf && cf-agent -K
$ rm /tmp/hello ; cf-agent -KI
info: Created file '/tmp/hello', mode 0600
info: Updated content of '/tmp/hello' with content 'cfengine'
In this example, control_common_bundlesequence_end
is a special variable, handled by the Masterfiles Policy Framework (MPF).
To learn about more variables like this and ways to interact with the MPF without editing it, see the [MPF Reference documentation][Masterfiles Policy Framework].
The rest of this documentation page below focuses on the specifics of how augments files work, independently of everything they can be used for in the MPF.
There are two canonical augments files, host_specific.json
, and def.json
which may load additional Augments
as specified by the [augments key][Augments#augments].
Notes:
- CFEngine variables are not expanded unless otherwise noted.
If $(sys.workdir)/data/host_specific.json
(typically /var/cfengine/data/host_specific.json
) is the first augments file that is loaded. Any variables defined as a result of processing this file
are automatically tagged with source=cmdb
. Variables defined from this file can not be overridden by subsequently processed augments files. Policy always wins and thus can overwrite the variable.
Notes:
- This file does not support the [augments key][Augments#augments].
The file def.json
is found based on the location of the policy entry (the first policy file read by the agent):
- with no arguments, it's in
$(sys.inputdir)/def.json
because$(sys.inputdir)/promises.cf
is used - with
-f /dirname/myfile.cf
, it's in/dirname/def.json
- with
-f ./myfile.cf
, it's in./def.json
Notes:
sys
variables are expanded indef.json
and all subsequently loaded augments as specified by theaugments
key.def_preferred.json
will be used instead ofdef.json
if it is present. This preferential loading can be disabled by providing the--ignore-preferred-augments
option to the agent.
An augments file can contain the following keys:
This key is supported in def.json
, def_preferred.json
, and augments loaded by the [augments key][Augments#augments].
Filenames entered here will appear in the def.augments_inputs
variable.
Notes:
-
Files are loaded relative to
sys.policy_entry_dirname
. -
The inputs key has precedence over the vars key.
-
If both the inputs key and
vars.augments_inputs
are populated concurrently, the variabledef.augments_inputs
will hold the value set by the inputs key. Thedef.augments_inputs
variable is part of the default inputs in theMasterfiles Policy Framework
.
Examples:
{
"inputs": [ "services/hello-world.cf", "example.cf", "/tmp/my_policy.cf" ],
"vars": {
"augments_inputs": [ "goodbye.cf" ]
}
}
The above Augments results in $(sys.policy_entry_dirname)/services/hello-world.cf
, $(sys.policy_entry_dirname)/example.cf
and /tmp/my_policy.cf
being added to inputs.
{
"vars": {
"augments_inputs": [ "goodbye.cf" ]
}
}
The above Augments results in $(sys.policy_entry_dirname)/goodbye.cf
being added to inputs.
This key is supported in both host_specific.json
, def.json
, def_preferred.json
, and augments loaded by the [augments key][Augments#augments].
Variables defined here can target a namespace and or bundle scope explicitly. When defined from host_specific.json
, variables default to the variables
bundle in the data
namespace ($(data:variables.MyVariable)
).
For example:
{
"variables": {
"VariableWithImplicitNamespaceAndBundle": {
"value": "value"
}
}
}
Variables can target the implicit namespace while specifying the bundle.
For example:
{
"variables": {
"my_bundle.VariableWithImplicitNamespace": {
"value": "value"
}
}
}
Variables can also target a namespace explicitly.
For example:
{
"variables": {
"MyNamespace:my_bundle.Variable": {
"value": "value"
}
}
}
The comment
key is optional, if supplied, the comment will be associated with the variable definition as if you had applied the comment attribute to a vars type promise.
For example, this JSON:
{
"variables": {
"MyNamespace:my_bundle.Variable": {
"value": "value",
"comment": "An optional note about why this variable is important"
}
}
}
Is equivalent to this policy:
body file control
{
namespace => "MyNamespace";
}
bundle agent my_bundle
{
vars:
"Variable"
string => "value",
comment => "An optional note about why this variable is important";
}
The tags
key is optional, if supplied, the tags will be associated with the variable definition as if you had applied the meta attribute to a vars type promise.
For example, this JSON:
{
"variables": {
"MyNamespace:my_bundle.Variable": {
"value": "value",
"tags": [ "inventory", "attribute_name=My Inventory" ]
}
}
}
Is equivalent to this policy:
body file control
{
namespace => "MyNamespace";
}
bundle agent my_bundle
{
vars:
"Variable"
string => "value",
meta => { "inventory", "attribute_name=My Inventory" };
}
Notes:
vars
andvariables
keys are allowed concurrently in the same file.- If
vars
andvariables
keys in the same augments file define the same variable, the definition provided by thevariables
key wins.
History:
- Added in 3.18.0
This key is supported in both host_specific.json
, def.json
, and def_preferred.json
and augments loaded by the augments key.
Variables defined here can target a namespace and or bundle scope
explicitly. When defined from def.json
, variables default to the def
bundle in the default
namespace ($(default:def.MyVariable)
).
Thus:
{
"vars": {
"phone": "22-333-4444",
"myplatform": "$(sys.os)",
"MyBundle.MyVariable": "MyValue in MyBundle.MyVariable",
"MyNamespace:MyBundle.MyVariable": "MyValue in MyNamespace:MyBundle.MyVariable"
}
}
results in the variable default:def.phone
with value 22-333-4444
,
default:def.myplatform
with the value of your current OS,
default:MyBundle.MyVariable
with the value MyValue in MyBundle.MyVariable
and MyNamespace:MyBundle.MyVariable
with the value MyValue
in
MyNamespace:MyBundle.MyVariable
.
Again, note that this happens before policy is parsed or evaluated.
You can see the list of variables thus defined in the output of cf-promises --show-vars
(see [Components][]). They will be tagged with
the tag source=augments_file
. For instance, the above two variables (assuming
you placed the data in $(sys.inputdir)/def.json
) result in
cf-promises --show-vars=default:def
...
default:def.myplatform linux source=augments_file
default:def.phone 22-333-4444 source=augments_file
Variables of other types than string can be defined too, like in this example
"vars" : {
"str1" : "string 1",
"num1" : 5,
"num2" : 3.5
"slist1" : ["sliststr1", "sliststr2"]
"array1" : {
"idx1" : "val1",
"idx2" : "val2"
}
}
Notes:
vars
andvariables
keys are allowed concurrently in the same file.- If
vars
andvariables
keys in the same augments file define the same variable, the definition provided by thevariables
key wins.
History:
- 3.18.0 gained ability to specify the namespace and bundle the variable should be defined in.
This key is supported in both host_specific.json
, def.json
, def_preferred.json
, and augments loaded by the augments key.
Any class defined via augments will be evaluated and installed as [soft classes][Classes and decisions]. This key supports both array and dict formats.
For an array each element of the array is tested against currently defined
classes as an [anchored regular expression][anchored] unless the string ends with ::
indicating it should be interpreted as a
[class expression][Classes and decisions].
For example:
{
"classes": {
"augments_class_from_regex_my_always": [ "any" ],
"augments_class_from_regex_my_other_apache": [ "server[34]", "debian.*" ],
"augments_class_from_regex_my_other_always": [ "augments_class_from_regex_my_always" ],
"augments_class_from_regex_when_MISSING_not_defined": [ "^(?!MISSING).*" ],
"augments_class_from_regex": [ "cfengine_\\d+" ],
"augments_class_from_single_class_as_regex": [ "cfengine" ],
"augments_class_from_single_class_as_expression": [ "cfengine::" ],
"augments_class_from_classexpression_and": [ "cfengine.cfengine_3::" ],
"augments_class_from_classexpression_not": [ "!MISSING::" ],
"augments_class_from_classexpression_or": [ "cfengine|cfengine_3::" ],
"augments_class_from_classexpression_complex": [ "(cfengine|cfengine_3).!MISSING::" ]
}
}
The tags
, comment
, and the mutually exclusive class_expressions
, and regular_expressions
subkeys
are supported when using the dict structure.
For example:
{
"classes": {
"myclass_defined_by_augments_in_def_json_3_18_0_v0": {
"class_expressions": [ "linux.redhat::", "cfengine|linux::" ],
"comment": "Optional description about why this class is important",
"tags": [ "optional", "tags" ]
},
"myclass_defined_by_augments_in_def_json_3_18_0_v1": {
"regular_expressions": [ "linux.*", "cfengine.*" ],
"tags": [ "optional", "tags" ]
}
}
}
Note that augments is processed at the very beginning of agent evaluation. You can use any hard classes, [persistent classes][Classes and decisions] , or classes defined earlier in the augments list. Test carefully, custom [soft classes][Classes and decisions] may not be defined early enough for use. Thus:
{
"classes": {
"augments_class_from_regex_my_always": [ "any" ],
"augments_class_from_regex_my_other_apache": [ "server[34]", "debian.*" ],
"augments_class_from_regex_my_other_always": [ "augments_class_from_regex_my_always" ],
"augments_class_from_regex_when_MISSING_not_defined": [ "^(?!MISSING).*" ],
"augments_class_from_regex": [ "cfengine_\\d+" ],
"augments_class_from_single_class_as_regex": [ "cfengine" ],
"augments_class_from_single_class_as_expression": [ "cfengine::" ],
"augments_class_from_classexpression_and": [ "cfengine.cfengine_3::" ],
"augments_class_from_classexpression_not": [ "!MISSING::" ],
"augments_class_from_classexpression_or": [ "cfengine|cfengine_3::" ],
"augments_class_from_classexpression_complex": [ "(cfengine|cfengine_3).!MISSING::" ],
"myclass_defined_by_augments_in_def_json_3_18_0_v0": {
"class_expressions": [ "linux.redhat::", "cfengine|linux::" ],
"comment": "Optional description about why this class is important",
"tags": [ "optional", "tags" ]
},
"myclass_defined_by_augments_in_def_json_3_18_0_v1": {
"regular_expressions": [ "linux.*", "cfengine.*" ],
"tags": [ "optional", "tags" ]
}
}
}
results in
-
augments_class_from_rgex_my_always
being always defined. -
augments_class_from_regex_my_other_apache
will be defined if the classesserver3
orserver4
are defined, or if any class starting withdebian
is defined. -
augments_class_from_regex_my_other_always
will be defined becauseaugments_class_from_regex_my_always
is listed first and always defined. -
augments_class_from_regex_when_MISSING_not_defined
will be defined if the classMISSING
is not defined. -
augments_class_from_single_class_as_regex
will be defined because the classcfengine
is always defined. -
augments_class_from_single_class_as_expression
will be defined becausecfengine
is defined when interpreted as a class expression. -
augments_class_from_classexpression_and
will be defined because the classcfengine
and the classcfengine_3
are defined and the class expressioncfengine.cfengine_3::
evaluates to true. -
augments_class_from_classexpression_not
will be defined because the class expression!MISSING::
evaluates to false since the classMISSING
is not defined. -
augments_class_from_classexpression_or
will be defined because the class expressioncfengine|cfengine_3::
evaluates to true since at least one ofcfengine
orcfengine_3
will always be defined by cfengine 3 agents. -
augments_class_from_classexpression_complex
will be defined because the class expression(cfengine|cfengine_3).!MISSING::
evaluates to true since at least one ofcfengine
orcfengine_3
will always be defined by cfengine 3 agents andMISSING
is not defined. -
myclass_defined_by_augments_in_def_json_3_18_0_v0
will be defined because the class expressioncfengine|linux::
will always be true since there is always acfengine
class defined. -
myclass_defined_by_augments_in_def_json_3_18_0_v1
will be defined because the expressioncfengine.**
will match at least one defined class,cfengine
You can see the list of classes thus defined through def.json
in the output
of cf-promises --show-classes
(see [Components][]). They
will be tagged with the tags source=augments_file
. For instance:
% cf-promises --show-classes=my
Class name Meta tags Comment
augments_class_from_regex_my_always source=augments_file
augments_class_from_regex_my_other_always source=augments_file
augments_class_from_regex_my_other_apache source=augments_file
myclass_defined_by_augments_in_def_json_3_18_0_v0 optional,tags,source=augments_file Optional description about why this class is important
myclass_defined_by_augments_in_def_json_3_18_0_v1 optional,tags,source=augments_file
See also:
- Functions that use regular expressions with classes:
classesmatching()
,classmatch()
,countclassesmatching()
History:
- 3.18.0
- Support for dict structure for classes and support for metadata (
comment
,tags
) added. - Classes are defined as soft classes instead of hard classes.
- Support for dict structure for classes and support for metadata (
This key is supported in def.json
, def_preferred.json
, and augments loaded by the [augments key][Augments#augments].
A list of file names that should be merged using mergedata()
semantic
Example:
Here we merge a platform specific augments on to the def.json
loaded next to
the policy entry and see what the resulting variable values will
be.
The def.json
next to the policy entry:
{
"vars":{
"my_var": "defined in def.json",
"my_other_var": "Defined ONLY in def.json"
},
"augments": [
"/var/cfengine/augments/$(sys.flavor).json"
]
}
The platform specific augments on a CentOS 6 host:
/var/cfengine/augments/centos_6.json
:
{
"vars": {
"my_var": "Overridden in centos_6.json",
"centos_6_var": "Defined ONLY in centos_6.json"
}
}
The expected values of the variables defined in the def bundle scope:
R: def.my_var == Overridden in centos_6.json
R: def.my_other_var == Defined ONLY in def.json
R: def.centos_6_var == Defined ONLY in centos_6.json
- 3.18.0
- Introduced
variables
key with support for metadata (comment
,tags
) and targeting the namespace and bundle. - Introduced ability for
vars
to target namespace and bundlevariables
key with support for metadata (comment
,tags
). - Introduced metadata (
comment
,tags
) support forclasses
key. - Introduced
def_preferred.json
and--ignore-preferred-augments
to disable it. - Classes defined from augments are now soft classes and not hard classes.
- Introduced parsing of
$(sys.workdir)/data/host_specific.json
- Introduced
- 3.12.2, 3.14.0 introduced class expression interpretation (
::
suffix) to classes key - 3.12.0 introduced the
augments
key - 3.7.3 back port
def.json
parsing in core agent and loaddef.json
if present next to policy entry - 3.8.2 removed core support for
inputs
key, loaddef.json
if present next to policy entry - 3.8.1
def.json
parsing moved from policy to core agent for resolution of classes and variables to be able to affect control bodies - 3.7.0 introduced augments concept into the Masterfiles Policy Framework