Distributed Monitoring System ganglia configuration document

Components of a ganglia

Ganglia includes the following program, between them through the XML format to pass monitoring data, to achieve monitoring effect.

Gmetad This program collects the data for each cluster and updates it to the RRD database.

Gmond collects the monitoring data of this machine and sends it to other machines. Collect monitoring data from other machines for Gmetad to read. Default Listening Port 8649

Web front-end a web-based monitoring interface that needs to be installed with Gmetad on the same node, from Gmetad to fetch data, and to read RRD databases, generate pictures, and display them.

Second, the working mode of ganglia

Ganglia collection data work can be done in either unicast (unicast) or multicast (multicast) mode, which defaults to multicast mode.

Unicast: Send your own collected monitoring data to a specific one or several machines, you can cross the network segment

Multicast: Send your own collected monitoring data to all machines in the same network segment, while collecting the monitoring data sent by all the machines in the same network segment. Because it is sent in the form of a broadcast packet, it needs to be within the same network segment. But within the same network segment, you can define different sending channels.

Third, the installation of ganglia

Ganglia installation is very simple, divided into the monitoring management side (Gmetad) and client (Gmond) installation of two parts, has been written as a setup script, not described here.

The recommended installation version is ganglia-3.4.0, which can be downloaded and installed in http://sourceforge.net/projects/ganglia/files/.

Iv. Configuration of Ganglia

Here is a simple ganglia configuration architecture, a monitoring management end and multiple clients, we assume that Gmond works in multicast mode and that there is a Cluster1 cluster with 10 servers (from host1 to host10), within the same network segment.

1, monitoring management end configuration

Monitoring management side of the configuration file is gmetad.conf, installed in the ganglia etc directory can be found, this file configuration is very simple, described as follows:

Data_source "Cluster1" host1 host2

Xml_port 8651

Interactive_port 8652

Rrd_rootdir "/var/lib/ganglia/rrds"

The Data_source parameter defines the cluster name and the nodes in the cluster. Because of the multicast mode, each Gmond node has all the monitoring data of the node machine in the Cluster1, so it is not necessary to write all the nodes into the data_source. The proposed write is no less than 2, when the host1 node freezes, will automatically find HOST2 node to fetch data.

Xml_port defines an interactive port for collecting data summaries, which, if not specified, is 8651 by default, and can be obtained by telnet to the port to monitor all data collected by the client on the management side.

Interactive_port defines the port where the web-side gets the data, which is specified when configuring the Ganglia Web monitoring interface.

The Rrd_rootdir parameter defines the storage path of the RRD database, and Gmetad is updated to the corresponding RRD database in the directory after the monitoring data is collected.

The above definition of a monitoring cluster through the Data_source parameter, if you want to monitor multiple application systems, you can define the following ways:

Data_source "My cluster" localhost my.machine.edu:8649 1.2.3.5:8655

Data_source "My Grid" 1.3.4.7:8655 grid.org:8651 grid-backup.org:8651

Data_source "Another source" 1.3.4.7:8655 1.3.4.8

In the definition of cluster nodes, you can use the host name, IP address and other forms, can also add ports, if not port, the default is 8649 ports, and can set the frequency of data acquisition, such as the above "ten localhost, 50 1.3.4.7:8655" and so on, The data is collected every 10 seconds and 50 seconds respectively.