A plugin is a functional building block that extends Fluent Bit through defined interfaces. Inputs capture events, filters apply processing logic, and outputs deliver events onward. The same term also applies when functionality is compiled into the Fluent Bit executable rather than supplied as a separate deployable file. section 1.2 section 2.1.2
Plugin interfaces manage coupling and enable extensibility. They support source, target, and filtering capabilities, letting a pipeline combine independently selected components instead of requiring every integration to be built into application-specific logic. section 1.3.3 section 2.1.2
Choose plugins by the role they play in an event pipeline:
- Input plugins collect OS events, system metrics, application logs, event-driven data, or network-pulled data. section 3.1
- Filter plugins add custom processing logic; Lua is suited to quickly defining small filter behaviors. section 1.3.3
- Output plugins route processed events to destinations such as stdout or systems reachable through standard protocols. section 2.1.2 section 3.1
Inspect and compose plugins from the CLI
Use fluent-bit -i dummy -h to inspect configuration details for the dummy input plugin. Pipeline controls then name an input with -i, an output with -o, a filter with -F, and plugin properties with -p name=value; for example, stdout can be selected as an output. External plugins can be loaded by supplying their location with --plugin. section 2.1.2
Extensions can be precompiled in C, Go, or WebAssembly, while Lua offers a scripting path for filters. Go-based plugins may require C-to-Go type translation, regression testing across Fluent Bit revisions, compatible builds, and added deployment and support planning. section 1.3.3 section 9.5.2
Table 2.2 Log operation CLI controls
|
Short parameter
|
Long parameter
|
Description
|
|---|---|---|
-C
|
--custom
|
Enables the use of custom sources (inputs) or sinks (outputs).
|
-e
|
--plugin
|
Identifies where an external plugin should be loaded from.
|
-f
|
--flush
|
Allows us to change the flush frequency by providing an integer representing a number of seconds. (No value means that flush defaults to every second.) The flush represents the frequency at which the log events are processed downstream, regardless of how full the buffering capacity is.
|
-l
|
--log_file
|
Allows us to direct the log events to a log file rather than the console. The console still gets a brief startup configuration summary.
|
-w
|
--workdir
|
Defines where Fluent Bit puts any temporary files.
|
-H
|
--http
|
Enables the HTTP server, which can be used to address HTTP
GET REST calls that can be used to retrieve information about the Fluent Bit node, such as health, metrics, and build details. Except for the response to the root URL, the other endpoints start with
/api/v1/.
|
-P
|
--port
|
Allows a different port from the default
2020 to be defined for the HTTP server.
|
-s
|
--coro_stack_size
|
Sets the stack size for coroutines (defaults to 24 KB).
|
Use the log-operation controls to locate the external-plugin loading option and distinguish it from pipeline input, output, filter, and property controls. section 2.1.2
