MapR Control System Part 1: Dashboard and Setting Topology

Overview

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. The MCS provides job monitoring metrics that help you troubleshoot issues, such as which jobs required the most memory in a given week or which events caused job and task failures. Use the MCS to access, monitor, and perform administrative tasks on your cluster.

MapR Sandbox for Hadoop includes brief overviews and tutorials to help get you acquainted with some MCS features and functionality that you would use as a cluster administrator to ensure the cluster runs smoothly.

Use the tutorials to perform the following operations in the MCS:

Part 1 (This tutorial)

  • Explore the Dashboard view
  • Set up topology

Part 2 (Click here)

  • Create volumes
  • Take snapshots
  • Create mirror volumes

Part 3 (Click here)

  • Configure notifications and alarms
  • Review job metrics

MCS Dashboard

When you first login to the MCS, the Dashboard view appears by default. The Dashboard provides a summary of information about the cluster including a cluster heat map that displays the health of each node; an alarms summary; cluster utilization that shows the CPU, memory, and disk space usage; services running across the cluster; the number of available, unavailable, and under replicated volumes; MapReduce jobs.


Cluster Heat Map

The Cluster Heat Map is the first panel in the Dashboard view. It displays color coded squares that represent nodes in a cluster. The color of the square indicates node health. A green node indicates that a node is in good health, whereas red indicates that a node requires immediate attention.


Explore the heat map

  • Use the zoom slider in the tool bar to expand the node and see specific node details.

    You can use the filters in the tool bar to select which node details display. For example, you can filter by memory utilization and all nodes using 80% or more of their available memory display in the dashboard in red. You can also sort nodes by rack, name, or status.
  • Click on the node to see node details. A new view opens displaying the details. Expand and collapse the panels.

  • Click the Dashboard tab to return to the Dashboard view.
  • Click to view Heat Map controls and the node color legend.

    The color statuses and descriptions listed in the legend indicate what the health of the node would be if it changed from green to any of the colors shown.
  • Adjust the refresh rate to set how often the MCS refreshes the displayed cluster data.
  • Change the column count to set how many columns of nodes the MCS displays in a rack. If you have many nodes in a cluster and you want to see all the nodes in the cluster with a critical status, you can filter by the word 'critical' and the MCS displays all the nodes in the cluster that require immediate attention.
  • Click to close the Heat Map control panel.

Cluster Utilization, Jobs, Services, Volumes

In the panels to the right of the Heat Map, you can see cluster utilization, MapReduce jobs, services running across the cluster, as well as mounted and unmounted volumes in the cluster. You can click on multiple links in the panels for more details about the cluster and nodes.

Explore Cluster Utilization

  • View how much CPU, memory, and disk space the cluster is using.

Explore Jobs

  • Notice the links for running and queued jobs, and for blacklisted nodes.

  • Click any of these links. A new tab displays the JobTracker view. When you run MapReduce jobs, this view displays a list of running jobs and queued jobs, as well as nodes that get blacklisted. The system blacklists nodes when they do not heartbeat back to the master node for several seconds. The system assumes the node cannot perform tasks.

Explore Services

  • View the services running in the cluster.

  • Click any of the service links to see nodes running a particular service.

Explore Mounted and Unmounted Volumes

  • View the number of mounted and unmounted volumes in the cluster.

  • Click either of the volumes links. A new tab displays the Volumes view. The Volumes view displays a list of mounted or unmounted volumes with volume information, like where each volume is mounted and its physical topology.

Navigation Panel

The Navigation panel provides links to views that enable cluster monitoring, management, and configuration.

Explore the Navigation Panel

  • Click on any view in the Navigation panel and to explore options.

Topology

Topology is a description of the physical layout of the cluster hardware, so that MapR knows which nodes are on different racks. When your data is replicated, the copies go to separate racks. That way, if an entire rack goes down, you don't lose access to your data. Topology is a description of the locations of racks and nodes.

The diagram above shows a cluster with three racks, labeled "rack1", "rack2", and "rack3." Each rack has eight nodes, labeled "node1", "node2", "node3", and so on. To describe the cluster using the terms of physical topology, we use the following guidelines:

  • All the racks are inside the enclosing topology /data.
  • Each node is inside its rack.
  • Any given node can only be in one topology.

In the above example, the node "node2" on rack "rack1" would have the following topology:
/data/rack1/node2

Set Topology

  1. From the Dashboard view, click Nodes to see the Nodes view.

  2. Select the checkboxes for all the nodes on a single rack.
  3. Click Change Topology to set up a label that describes the rack.

Setting up the cluster's physical topology is an important step - not only does it help protect your data in case of a rack failure, it also enables data placement and job placement features. For more information, see Node Topology.

Tutorial Category Reference: