The MapR Control System (MCS) is a graphical, programmatic control panel for cluster administration that provides complete cluster monitoring functionality and most of the functionality of the command line. This is Part 3 of the three-part series on MCS tutorials that talks about setting up Alarms and Notifications and the Metrics available on MCS.
Use the tutorials to perform the following operations in the MCS:
Part 1 (Click here)
- Explore the Dashboard view
- Set up topology
Part 2 (Click here)
- Create volumes
- Take snapshots
- Create mirror volumes
Part 3 (This tutorial)
- Configure notifications and alarms
- Review job metrics
Alarms and Notifications
Alarms alert you to issues within the cluster. Alarms provide information about overall cluster health, including disk failures. Alarms tell you which volumes are under-replicated or have reached quota, as well as when services stop running on a node.
MapR automatically raises node, volume, and user and group alarms. When the system raises an alarm, it can send email notifications to cluster, node, or volume owners. You can configure email addresses for alarms to notify additional people each time the system raises a new alarm.
Explore alarms in the MCS:
- In the Navigation pane, expand the Cluster group and click the Dashboard view. Alarms display in the Alarms pane of the Dashboard.
- Optionally, you can expand the Alarms group and click either the Volume Alarms view or the Node Alarms view.
Configure an alarm notification:
- In the Navigation pane, expand the Alarms group and click the Alerts view.
- Click Alarm Notifications to display the Alerts dialog.
- Select the Standard Notification checkbox to send notifications to the owner of the cluster, node, volume, or entity.
- Enter email addresses in the Additional Email Address fields to send alarm notifications to additional people.
- Click OK to save the settings.
When you install MapR, you can configure a MySQL database to store metrics. The MapR Metrics service collects and displays metrics for MapReduce jobs, tasks, and task attempts that run on the nodes in your cluster. The metrics help predict cluster usage, measure which jobs use the most resources, and troubleshoot the root causes of failures or performance problems.
You can view the following metrics for jobs, tasks, and task attempts:
- Cumulative CPU and memory usage
- Number of running or failed tasks or attempts
- Speed of input, output, and shuffle
- Duration of task attempts
- Amount of data read, written, or shuffled
- Amount of memory currently in use
- Number of records skipped or spilled
Note: The MapR Sandbox for Hadoop does not contain metrics data.
View job metrics:
- In the Navigation pane, expand the Cluster group and click the Jobs view. A histogram displays the distribution of jobs by job duration, as well as a list of all jobs that have run on the cluster.
- Hover the cursor over a bar in the histogram and click Filter to filter results down to that particular job.
- Click on the job name in the list to drill down into the job. A new view for the job appears and displays all job tasks.
- Click on a task ID in the list to drill down into the task and view task attempts.
- Select the Chart and Task views to see the variation between the two views.
In this three-part series on MCS, you have learned how to navigate the MapR Control System interface, as well as how to perform some administrative tasks on a cluster. You set up rack topology, created a volume, snapshot, and mirrors, configured a notification, and viewed alarms and job metrics.