Simple Stat Generator
Sistagen (Simple Stat Generator) is console tool that simplifies graph plotting. It’s main job is to parse configuration and call rrdtool to generate desired graphs.
This app is kind of higher layer on rrdtool, allowing anything rrdtool allows, but in easier to manage form.
RRDTool itself is great tool, but it’s quite time consuming task to prepare graph with it directly. On the other side, there are many bigger or lesser (often all-in-one, and mostly web-oriented) applications, but they’re enforcing theirs philosophy, and having annoying “features”.
At first, take a look on graph definition:
echo "include templates.conf Graph test use disk-usage input-dir df-root" | sistagen
That would output file graph_400x100.png in current directory, containing disk space utilization graph. Of course, you have to have df-root/ directory with rrd files (generated with collectd - df plugin).
In this example, sistagen reads configuration from stdin. Thats because no input files was given. You can pass as many files as you want, just by giving they names of absolute path (TODO: describe more about pathes).
sistagen config1.conf config2.conf […]
Sistagen configuration consists of list of declarations and attributes.
Both graph and template means almost the same, with one difference: template isn’t graphed, it’s only ‘virtual’ declaration, while all graph declarations causes sistagen to plot them out.
Graph (and template) declarations should have some attributes. An attribute in sistagen configuration is line, starting with at last one whitespace and beginning with keyword, followed by number of named values. It’s important that attribute must always refer to declaration, so: 1. all attributes that follows a declaration, belongs to that declaration 2. it’s an error to pass attributes before any declaration
In its simplest case it looks like following:
GRAPH|TEMPLATE name attribute1 name=val ... attribute2 ... ...
Most of attributes should have named values (except of ‘use’, ‘merge’ and ‘size’). Each value should be in form: “name value” or “name=value” (the equation symbol is optional).
- use template-name - use template as current declaration base
- merge template-name [separator=S] - merge this declaration with the one given (see: difference between merge and use)
- colors theme-name - use theme-name as base for auto-colors
- title “Some title” - set graph title (same effect as: option title “Some title”)
- input name=NAME file=FRRDFILE ds=DS cf=CF - create new input named NAME from file RRDFILE using DS and CF There could be as many inputs as needed.
- output dir=PATH file=FILE prefix=PREFIX suffix=SUFFIX mask=MASK format=FORMAT When using output FILE then dir, mask and format are updated automatically, but prefix and suffix are cleared
- option option-name option-value
- option …
- … any option that may be passed into rrdtool graph, for example: lower-limit, zoom, start, etc..
- size SIZE [SIZE, …] any number of dimensions of graphs to be generated each size would be used in output file name. Each SIZE have to be in form: WxH where W - graph width, H - graph height
- date DATE [DATE, …] like SIZE above, allows to define several number of time periods, for which the graphs would be generated. See “Date period definition” below to learn more.
- line1|line2|line3|area (with EXPR | use NAME) [color COLOR] [text LEGEND] [stack]
- gprint text TEXT (with EXPR | use NAME) get AGGR currently supported graphing methods for lines and areas there should be at last one of ‘with’ or ‘use’ keyword with corresponding value EXPR means any rpn expression (http://oss.oetiker.ch/rrdtool/doc/rrdgraph_rpn.en.html) NAME means any input name defined earlier AGGR means an aggregation function (one of CF)
- comment text TEXT
Graph and template declarations may be inherited. It means that when one declaration inherits another then all attributes of parent would be copied into child.
Template A: attribute a Template B: use A # use A as base # attribue a is defined already
Use is generally used in the terms of inheritance. That means, it’ll cause to copy all attributes from ‘parent’. Also, ‘use’ will override any existing values. Merge on the other side, will NOT override any values, and will copy only input and graph declarations.
In summary: use is used in inheritance, merge is used in combining graphs.
Destination filename is created from several parts: OUTPUT-DIR + OUTPUT-PREFIX + OUTPUT-MASK + OUTPUT-SUFFIX + OUTPUT-FORMAT
OUTPUT-MASK may contain several keywords, that would be replaced during graph generation. They are:
- %(width)i - will be replaced with current graph width
- %(height)i - as above, but height
- %(name)s - current graph (declaration) name
- %(date_[from|to]_[full|date|time|year|month|day|hour|minute|second|days|hours|minutes|seconds])s the date of graphing period start (from) or end (to) which contains YYYYMMDDHHmmss (full) YYYYMMDD (date), HHmmss (time), or one of those components. days, hours, minutes and seconds meands time delta between given date and now
Default output mask looks like: _%(width)ix%(height)i
It is possible to define a “color theme” that defines colors per each graph element. Each one by default uses color=auto declaration, and in such situation sistagen pulls next available color from color theme (if number of graph elements outnumbers colors in color definition - random color is generated).
In general, color theme definition looks like
Theme THEME_NAME color value=RGBVAL [alpha=AA] [for=any|line|line123|area] color ... ...
Color defines new color definition, that would be used by graph elements when color=auto is set. It is possible to define any number of colors in theme declaration.
It’s possible to plot several graphs at once, each containing data from different periods. It is done by defining desired DATEs’ in graph definition.
Each DATE entry have form: FROM-TO The ‘TO’ part may be left, so it would be assigned as ‘now’. and each FROM and TO can be relative or absolute date definition. Relative date definitions are in form: NU - N - number, U - unit, for example: 10h, 5m, 3d, and so. Available units: s, m, h, d
Absolute date definitions are in form: [[YY]YYMMDD]HHmm[ss] year may be set as full year (2014) or shortened (14). The date part may be omitted, only time part left (HHmmss), also, seconds are not necessary.
There are some templates already prepared and shipped with applications, that lets fast and easy graph creation. Right now they are mostly focused on rrd files created by collectd, so with those files you may just create graph with only input-dir attribute.
Templates dir and current dir are added to config path (-C) by default.
Right now, all of the templates defined are prepared for stats generated by collectd daemon.
Here are some example graph definitions:
include templates.conf Graph cpu # this would graph cpu-0 usage use cpu input dir=/var/lib/collectd/rrd/localhost/cpu-0
include templates.conf Graph load # cpu load average for 1, 5 and 15 min use load input dir=/var/lib/collectd/rrd/localhost/load
include templates.conf Graph bandwidth-eth0 # bandwidth utilization use bandwidth input dir=/var/lib/collectd/rrd/localhost/interface-eth0/ output prefix=bw-eth0
Template mybase # each graph in following dimensions size 400x50 1000x200 # plot 1day, 7days and one month date 1d 7d 30d include templates.conf Graph cpu # this would graph cpu-0 usage use cpu use mybase input dir=/var/lib/collectd/rrd/localhost/cpu-0
See “examples/” directory for more.
Please send me any reports and opinions as well. It would be nice to know if this tool is useful for someone, so I may still work on this. You’ll find my mail address in –help.