Pick the Best Open‑Source Monitoring Tool and Design a Unified Ops Platform

Comparison of Open‑Source Monitoring Tools

Cacti is a PHP‑based network‑traffic monitoring tool that uses SNMP (via snmpget and snmpwalk) to collect data, stores metrics with RRDTool, and renders graphs through a web interface.

Nagios is a free, open‑source monitoring system that watches hosts, services, switches, routers, printers, etc., and sends email or SMS alerts on failures and recoveries.

Zabbix provides distributed system and network monitoring via a web UI, supporting SNMP, agents, ping, and port checks, and runs on many Unix‑like platforms.

Ganglia is a scalable distributed monitoring system designed for HPC clusters. It consists of gmond daemons on each node, a gmetad aggregator, stores data with RRDTool, and presents it via a PHP front‑end.

Centreon builds on Nagios, using Nagios as the data collector while providing a web‑based configuration interface that simplifies Nagios setup and management.

Unified Operations Monitoring Platform Design

The platform centralizes monitoring of network, hardware, software, and database resources, standardizes data collection, storage, processing, presentation, authentication, and authorization, and aims for automated, intelligent operations management.

Ganglia Installation (YUM Method)

CentOS does not include Ganglia in the default repositories, so the EPEL repository must be added first. After enabling EPEL, install the two packages ganglia-gmetad and ganglia-gmond via yum.

Enable the EPEL repository (download and install the RPM).

Run yum install ganglia-gmetad ganglia-gmond.

Verify that /etc/ganglia contains the default configuration files.

After installation, gmetad runs on the monitoring server and gmond runs on each monitored client.

Ganglia Management‑Side Configuration

The main configuration file is gmetad.conf. Key parameters to edit include:

data_source : defines the cluster name and member nodes (e.g., Cluster1 cloud0 cloud2). At least two nodes are recommended for high availability.

gridname : name of the overall grid composed of multiple clusters.

xml_port : port for data aggregation (default 8651).

interactive_port : port used by the web front‑end to fetch data.

rrd_rootdir : directory where RRD files are stored.

Ganglia Client Configuration

Each client runs gmond with its configuration in /etc/ganglia/gmond.conf. The file defines the host name, cluster name, and the list of multicast or unicast peers.

Web Front‑End Setup

The Ganglia web UI is PHP‑based. After installing a PHP environment, download the latest ganglia-web package (e.g., version 3.7.1) from SourceForge and place it under the Apache document root.

Rename conf_default.php to conf.php (or keep the default) and adjust the following settings: $conf['dwoo_compiled_dir'] and $conf['dwoo_cache_dir'] must exist and be writable (chmod 777).

Ensure the RRD directory (e.g., /opt/app/ganglia/rrds) is writable by the web user.

Extending Ganglia

Two common extension methods are demonstrated:

gmetric : a command‑line tool that sends custom metrics to a gmond node. Example options include -n (metric name), -v (value), -t (type), -u (unit), -d (lifetime), -c (config file), and -S (spoof source).

Python plugins : ready‑made modules are available at https://github.com/ganglia/gmond_python_modules.

Advantages and Caveats

Scales to monitor tens of thousands of servers with data latency under 10 seconds.

Distributed architecture suits multi‑site, cross‑datacenter deployments.

Integrates seamlessly with Centreon/Nagios for unified alerting.

Disk I/O for RRD storage can become a bottleneck; high‑performance disks are recommended.