Cacti collects measurements, stores them in Round Robin Database (RRD) files, and uses RRDtool to render graphs. A MySQL or MariaDB database holds device configuration, templates, permissions, polling state, and intermediate results. RRD files hold the time-series measurements.
A Device represents something Cacti monitors, such as a server, switch, sensor, or service. Its configuration identifies the responsible data collector, address, availability checks, and SNMP settings where needed. Site assignments organize devices by location; they do not create separate database or security boundaries.
Device Templates associate graph templates and data queries with a device. Automation can discover devices and apply rules for creating devices, graphs, and tree entries.
The main Data Collector runs on the central Cacti system. Remote collectors poll devices closer to their location and exchange configuration and results with the main system. During a loss of connectivity, a configured remote collector can retain results locally for later delivery. This does not replace backups or provide automatic failover of the main server.
An operating-system scheduler or the Cacti daemon starts poller.php.
It coordinates collection and launches the configured engine: the PHP
cmd.php poller or the C-based Spine poller. The scheduler
interval,
poller interval, and data-source profiles must agree; changing cron alone does
not change an existing RRD file's sampling interval.
Cacti uses SNMP, external scripts, and resident PHP script-server functions. The available SNMP algorithms depend on the device, Net-SNMP, PHP, and Spine builds. Scripts execute with the poller's operating-system permissions, not with an interactive administrator's environment.
Data Input Methods describe commands and their input and output fields. Data Queries enumerate indexed objects, such as network interfaces or disk partitions. Query input fields describe those objects; output fields provide measurements to collect.
Data Sources connect collection definitions to RRD files and the series stored in them. The poller writes results to database output tables. Cacti's update processing writes those measurements to RRD files, either during normal polling or through Boost when on-demand updating is enabled. Spine does not directly render graphs or replace the RRD update process.
Data Source Profiles define sampling, heartbeat, consolidation functions, and retention when files are created. RRD archives have bounded storage for their configured layout. Older measurements are consolidated or expire according to that layout; changing a graph's timespan does not recover detail already consolidated away.
An RRD file can contain multiple named series. Database data-source counts, poller item counts, graph counts, and RRD file counts therefore need not match.
Graph Templates define presentation settings and graph items. A graph item can draw a series, calculate a value, or add legend text. Several graphs can reuse an existing data source without collecting it again. CDEFs, VDEFs, and GPRINT presets control calculations and formatting at graph-rendering time.
Users view permitted graphs through the tree, list, and preview views described in Viewing Graphs. Permissions and tree publication control which graphs a user can reach; adding a graph to a tree does not start polling.
Plugins add features through Cacti's plugin API. Check each plugin's release compatibility, installation instructions, and permissions before enabling it. Template import and export share definitions; they do not transfer historical RRD measurements. See Plugin Management and Template Import.
Copyright (c) 2004-2026 The Cacti Group