An Object-Oriented Python Interface for RRDTool
PyRRD lets you use RRDTool from Python code that takes advantage of standard object-oriented patterns. The makes the programmatic usage of RRDTool much easier and reusable.
A quick review of features is available at the project wiki  . Example code with graph image output is also available on the wiki  .
PyRRD is an object-oriented wrapper for the command line graphing and round-robin database utility, rrdtool  . PyRRD originally had two design goals:
provide an interface to rrdtool that Python programmers would love, and
not depend upon the Python bindings for rrdtool.
The reasons for the former are obvious. The motivation for the latter were the many people who had difficulty compiling the rrdtool bindings on their operating system of choice.
Even though PyRRD’s original purpose was to help those without the bingins, the project now offers support for those with the bindings installed. As such, users may enjoy both the speed benefits from the bindings as well as the API usability from PyRRD.
For docs, see the docstrings at the beginning of each class (and many of the functions). They not only contain many of the standard RRDTool docs, but they contain doctests which give you a hands-on, how-it-works understanding of actual usage.
For those curious about the motivation for creating PyRRD, perhaps some background would be in order. Originally, there were only two ways to use RRDTool from Python:
using the Python bindings which were difficult to compile and use, or
2. making system calls to the “rrdtool” executable from Python, passing it all the parameters it needed.
Option #1 was often difficult or impossible for many folks to get running on their preferred operating system. But even if one was able to compile it and run it, usage was very cumbersome and designed to work like the command line tool and the C interface, not like most people typically use Python.
Now, with PyRRD, there are two additional ways to use RRDTool from Python:
3. an object-oriented interface that wraps system calls (Popen) to the rrdtool binary, and
4. the same object-oriented interface that wraps the cumbersome rrdtool Python bindings.
Some parts of PyRRD make use of ElementTree for XML processing. If you have Python 2.5 or greater, PyRRD will use xml.etree. If your Python version is less than 2.5 and you want to use features that depend on XML processing (such as dump function and the fetch/info methods), you will need to install the ElementTree library  .
PyRRD is installed in the usual way:
python setup.py install
You may also use PyRRD without installing it as long as you have ./ in your PYTHONPATH and you are in the top-level directory (which has the pyrrd child directory).
Create an RRD file programmatically:
>>> from pyrrd.rrd import DataSource, RRA, RRD >>> filename = '/tmp/test.rrd' >>> dataSources =  >>> roundRobinArchives =  >>> dataSource = DataSource( ... dsName='speed', dsType='COUNTER', heartbeat=600) >>> dataSources.append(dataSource) >>> roundRobinArchives.append(RRA(cf='AVERAGE', xff=0.5, steps=1, rows=24)) >>> roundRobinArchives.append(RRA(cf='AVERAGE', xff=0.5, steps=6, rows=10)) >>> myRRD = RRD( ... filename, ds=dataSources, rra=roundRobinArchives, start=920804400) >>> myRRD.create()
Let’s check to see that the file exists:
>>> import os >>> os.path.isfile(filename) True
Let’s see how big it is (depending upon RRDTool version, the byte count can change, so we’ll just get a general sense):
>>> bytes = len(open(filename).read()) >>> 800 < bytes < 1200 True
In order to save writes to disk, PyRRD buffers values and then writes the values to the RRD file at one go:
>>> myRRD.bufferValue('920805600', '12363') >>> myRRD.bufferValue('920805900', '12363') >>> myRRD.bufferValue('920806200', '12373') >>> myRRD.bufferValue('920806500', '12383') >>> myRRD.update()
Let’s add some more data:
>>> myRRD.bufferValue('920806800', '12393') >>> myRRD.bufferValue('920807100', '12399') >>> myRRD.bufferValue('920807400', '12405') >>> myRRD.bufferValue('920807700', '12411') >>> myRRD.bufferValue('920808000', '12415') >>> myRRD.bufferValue('920808300', '12420') >>> myRRD.bufferValue('920808600', '12422') >>> myRRD.bufferValue('920808900', '12423') >>> myRRD.update()
If you’re curious, you can take a look at your rrd file with the following:
The output of that isn’t printed here, ‘cause it take up too much space. However, it is very similar to the output of the similarly named rrdtool command.
In order to create a graph, we’ll need some data definitions. We’ll also throw in some calculated definitions and variable definitions for good meansure:
>>> from pyrrd.graph import DEF, CDEF, VDEF, LINE, AREA, GPRINT >>> def1 = DEF(rrdfile=myRRD.filename, vname='myspeed', ... dsName=dataSource.name) >>> cdef1 = CDEF(vname='kmh', rpn='%s,3600,*' % def1.vname) >>> cdef2 = CDEF(vname='fast', rpn='kmh,100,GT,kmh,0,IF') >>> cdef3 = CDEF(vname='good', rpn='kmh,100,GT,0,kmh,IF') >>> vdef1 = VDEF(vname='mymax', rpn='%s,MAXIMUM' % def1.vname) >>> vdef2 = VDEF(vname='myavg', rpn='%s,AVERAGE' % def1.vname) >>> line1 = LINE(value=100, color='#990000', legend='Maximum Allowed') >>> area1 = AREA(defObj=cdef3, color='#006600', legend='Good Speed') >>> area2 = AREA(defObj=cdef2, color='#CC6633', legend='Too Fast') >>> line2 = LINE(defObj=vdef2, color='#000099', legend='My Average', ... stack=True) >>> gprint1 = GPRINT(vdef2, '%6.2lf kph')
Color is the spice of life. Let’s spice it up a little:
>>> from pyrrd.graph import ColorAttributes >>> ca = ColorAttributes() >>> ca.back = '#333333' >>> ca.canvas = '#333333' >>> ca.shadea = '#000000' >>> ca.shadeb = '#111111' >>> ca.mgrid = '#CCCCCC' >>> ca.axis = '#FFFFFF' >>> ca.frame = '#AAAAAA' >>> ca.font = '#FFFFFF' >>> ca.arrow = '#FFFFFF'
Now we can create a graph for the data in our RRD file:
>>> from pyrrd.graph import Graph >>> graphfile = "/tmp/rrdgraph.png" >>> g = Graph(graphfile, start=920805000, end=920810000, ... vertical_label='km/h', color=ca) >>> g.data.extend([def1, cdef1, cdef2, cdef3, vdef1, vdef2, line1, area1, ... area2, line2, gprint1]) >>> g.write()
Let’s make sure it’s there:
>>> os.path.isfile(graphfile) True
Let’s get a sense of the byte size:
>>> bytes = len(open(graphfile).read()) >>> bytes != 0 True >>> 8000 < bytes < 10400 True
Open that up in your favorite image browser and confirm that the appropriate RRD graph is generated.
Let’s clean up the files we’ve put in the temp directory:
>>> os.unlink(filename) >>> os.unlink(graphfile)
In addition to the command line tool, PyRRD also supports using the Python bindings, if you have them installed. Let’s set some stuff up like we did in the previous example:
>>> filename = '/tmp/test.rrd' >>> dataSources =  >>> roundRobinArchives =  >>> dataSource = DataSource( ... dsName='speed', dsType='COUNTER', heartbeat=600) >>> dataSources.append(dataSource) >>> roundRobinArchives.append(RRA(cf='AVERAGE', xff=0.5, steps=1, rows=24)) >>> roundRobinArchives.append(RRA(cf='AVERAGE', xff=0.5, steps=6, rows=10))
Usage is identical to the standard PyRRD usage, with the exception of object construction:
>>> from pyrrd.backend import bindings >>> myRRD = RRD(filename, ds=dataSources, rra=roundRobinArchives, ... backend=bindings) >>> myRRD.create()
Note that since the Graph module is its own beast, you will need to indicate whether you want to use the bindings or the external backend when you graph as well:
>>> from pyrrd.graph import Graph >>> graphfile = "/tmp/rrdgraph.png" >>> g = Graph(graphfile, start=920805000, end=920810000, ... vertical_label='km/h', color=ca, backend=bindings) >>> g.data.extend([def1, cdef1, cdef2, cdef3, vdef1, vdef2, line1, area1, ... area2, line2, gprint1]) >>> g.write()
Everything else is the same. Let’s check to see that the file exists, and then we’ll cleanup:
>>> import os >>> os.path.isfile(filename) True >>> os.unlink(filename)
Move test code around (testing and admin/testRunner).
Fix breaking tests.
Add wiki examples for using info and fetch
Improve the wrapper for the Python RRDTool bindings
a lot of the code in PyRRD was written quite a while ago (circa 2004), and needs to be refactored (removing redundancies, use better idioms, etc.)
Allow for users to supply their own fd to pyrrd.graph.
Update all examples for recent dates like example4 has been updated.
Stop using actual file writes and doctests for file tests; use unit tests (and StringIO) instead.
Add an RPN class.
Add a DS collection class that has a get() method for getting a particular DS by name.
Add support for atomic operations.
From 0.0.7 to 0.1.0
This version marks a significant update to the code base. Some 70+ commits have been made to trunk since the last release in the period from March 2009 to September 2011. Some of the most signficant changes include the following:
Added a wrapper for the Python bindings.
Improvements in tests and testing infrastructure.
Improved exception handling.
Many changes and improvements to the RRD -> Python object mapper.
Improved support in Windows.
Graph formatting fixes.
Many bug fixes from community members.
From 0.0.6 to 0.0.7
Packaging improvements and loads of documentation.
From 0.0.5 to 0.0.6
Bug fix release (missing files in source package).
From 0.0.4 to 0.0.5
Added support for retrieving and displaying RRD from RRD files.
Added an object mapper for RRD data (via XML files).
Added community-contributed improvements.
From 0.0.3 to 0.0.4
Updated all the examples to work with the latest code.
Added community-contributed bug fix for Windows users.
From 0.0.2 to 0.0.3
Minor code reorg.
Various bug fixes.
From 0.0.1 to 0.0.2
Added unit tests.
Added more examples.
From 0 to 0.0.1
Reorganized RRD code as donated from the CoyMon project.
Got basic rrdtool functionality represented as Python classes.
The following members of the community have provided valuable contributions to this project:
Ravi Bhalotia, Allen Lerner, Mike Carrick and the U.S. Department of Veterans Affairs
AdytumSolutions, Inc., E-Secure Systems
nasvos, Leem Smit, Aaron Westendorf and Agora Games
Mladen Milankovic, Denis Fortin
Joseph Heck, Jean-Baptiste Quenot
Pavel Shramov, Brad Beattie, Colin Horsington
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.