Configuring Probes for the Autonomic

Automatic reactions designate the actions undertaken by the DM (the autonomic engine to be precise) when events and alarms are emitted by agents. What is done when conditions are met must be defined in a Roboconf commands file.

These files use a sub-set of the Drools’s syntax.
Single-line comments can start with # or with // (C-style).
Multi-line comments are enclosed between /* and */.

The global structure of a rule is…

rule "name"
	attributes
	when
		event name
	then
		list of command files
end

Rule files must be located under the rules.autonomic directory of the application (template)’s directory. One rule, one file. The rule name must match the file name. the file extension is .drl (the same than Drools rules).

Therefore, the following rule should be defined…

rule "scale"
	when
		cpuIsTooHigh
	then
		replicateMachine
end

… in the scale.drl file.

Examples

The following rule indicates that when the cpuIsToHigh event arises, the replicateMachine command file will be executed.

rule "basic-scale"
	when
		cpuIsTooHigh
	then
		replicateMachine
end

It is possible to execute (sequentially) several command files.

rule "stronger-scale"
	when
		cpuIsTooHigh
	then
		replicateMachine;
		replicateThis
		replicateThatToo
end

Commands refer to the name of a commands file (without the extension).
They can be separated by a colon or by a new line.

Attribute: sleep period is

This attribute allows to define a delay between which this rule will not be fired again. A rephrasing can go as the following scenario:

  1. A combination of events results in the activation of a rule called “r”.
  2. This rule defines its sleep period as 120 seconds.
  3. 30 seconds later, the combination of triggering events appears again.
    But since the rule was already activated and that the sleep delay is not over, the rule will not be executed.
  4. Five minutes later, the right combination is again appearing. Since the sleep period is over, the rule can be executed again.

The sleep period can be seen as the period necessary for the rule (and the associated commands) to produce the required effects. Such a period is defined as follows…

rule "basic-scale"
sleep period is 50s
when
	cpuIsTooHigh
then
	replicateMachine
end

This attribute does not come from Drools’s ones.
It may be replaced later if a better name is found.

Notice the time unit is in seconds.
It is optional. It means you can write sleep period is 50s or sleep period is 50.

When not specified, there is no delay between execution.
Also, notice that a given combination of events only trigger a rule once. Then, only a new event can trigger this same rule.

Attribute: time window is

The time window is an anticipation of a future feature upgrade.

For the moment, a single event triggers the activation of a rule.
Later, it will be possible to define combination of events (this one and that one, etc). The time window is a period, measured in seconds, during which we consider the events that can trigger a rule.

As an example…

rule "basic-scale"
time window is 120s
when
	cpuIsTooHigh
	freeMemoryIsRunningOut
then
	replicateMachine
end

… implies that the cpuIsToHigh and freeMemoryIsRunningOut events must be both received during a period of 120 seconds. If they are separated by a period longer than the timing window, then the rule will not be activated.

This attribute does not come from Drools’s ones.
It may be replaced later if a better name is found.

Notice the time unit is in seconds.
It is optional. It means you can write time window is 120s or time window is 120.

When not specified, there is no timing window.
Any new event that triggers the rule will be considered.

Attribute: occurrences is

By default, a rule is activated when a given event is sent by an agent.
An event or a combination of events, that may come from a same or from different agents.

This attribute indicates that the event or the combination must occur a certain number of times. Knowing the delay between each agent check is for the moment hard-coded (20 seconds), it is possible to evaluate over time that a same problem or situation arises.

When occurrences is is used, then time windows is is required too.
It thus prevents old events from remaining considered. In the case of a combination of events, every time a combination is met, it is reset. If it occurs the expected number of times within the time window, then the rule will be activated.

This attribute does not come from Drools’s ones.
It may be replaced later if a better name is found.

The unit is a positive integer.
It means you can write occurrences is 5.

When not specified, the value is 1.
Any new event that triggers the rule will be considered.