Adding Alarms in Automation Console
You can add alarm defintions for any entity in the domain. See Defining Alarms. To add an alarm in the Automation Console:
- On the left side, click Dashboard.
- In the upper right corner of the Automation Console, select Add Widget > Alarm Definitions.
In the Alarm Definition widget, select Actions > Add to display the Add Alarm Definition dialog:
- After you enter a name and description for the alarm, specify the alarm type. This is either a discrete domain event, such as Node Joined (transaction engine or storage manager process started) or Node Left, or a statistics/metric alarm.
- The severity level indicates the importance of the alarm. The Automation Console shows Severe and Warning alarms but Info alarms are only logged.
- The alarm action indicates what happens when the alarm fires. Choose Notify to notify all management clients if this alarm fires. Choose Log Only to log the alarm in the broker's log file.
- Click Next to select the scope of the alarm definition, which can be domain, database, host or process.
If you select database, host or process, you then indicate whether the alarm definition applies to all entities or a particular one. For example:
As an exercise, you might want to add a node joined/left alarm definition. Then execute
killall nuodb on your host. The NuoDB enforcer for managed databases will restart the database, and alarms will then fire. The alarms will not show up immediately in your Automation Console because brokers observe events, queue alarms, and notify management clients periodically. After a few seconds, look at your Alarm widget, which might look something like this:
Metric alarms fire only after
n consecutive threshold breaches. Metrics are available for all database processes, databases (aggregating statistics of all of their TEs and SMs), hosts, as well as domain aggregation. For example, you can create an alarm that fires if the average CPU usage of all hosts in the domain is over 60% for at least 10 minutes:
In a metric alarm definition:
- Scope defines the entities that you are watching. You can watch domain (aggregates), database (aggregate of all engine nodes), host (OS statistics), and process.
- Aggregation applies if you are watching a database or domain metric. Indicate whether to use the sum or average.
- Breach Period indicates that an alarm fires only if consecutive metrics breach the threshold condition. The unit is in seconds. Brokers exchange metric information periodically with a significant delay so any breach period should be at least 60 seconds.
- The Threshold and Operator define the metric breach condition.