3 rrdtutorial - Alex van den Bogaerdt's RRDtool tutorial
5 =for html <div align="right">Go <a href="rrdtutorial.es.html">Spanish</a></div>
7 =for html <div align="right"><a href="rrdtutorial.pdf">PDF</a> version.</div>
11 RRDtool is written by Tobias Oetiker <oetiker@ee.ethz.ch> with
12 contributions from many people all around the world. This document is
13 written by Alex van den Bogaerdt <alex@ergens.op.het.net> to help you
14 understand what RRDtool is and what it can do for you.
16 The documentation provided with RRDtool can be too technical for some
17 people. This tutorial is here to help you understand the basics of
18 RRDtool. It should prepare you to read the documentation yourself.
19 It also explains the general things about statistics with a focus on
26 Please don't skip ahead in this document! The first part of this
27 document explains the basics and may be boring. But if you don't
28 understand the basics, the examples will not be as meaningful to you.
30 =head2 What is RRDtool ?
32 RRDtool refers to Round Robin Database tool.
33 Round robin is a technique that works with a fixed amount of data, and a
34 pointer to the current element. Think of a circle with some dots plotted
35 on the edge, these dots are the places where data can be stored. Draw an
36 arrow from the center of the circle to one of the dots, this is the pointer.
37 When the current data is read or written, the pointer moves to the next
38 element. As we are on a circle there is no beginning nor an end, you can
39 go on and on. After a while, all the available places will be used and
40 the process automatically reuses old locations. This way, the database
41 will not grow in size and therefore requires no maintenance.
42 RRDtool works with with Round Robin Databases (RRDs). It stores and retrieves
45 =head2 What data can be put into an RRD ?
47 You name it, it will probably fit. You should be able to measure some value
48 at several points in time and provide this information to RRDtool. If you
49 can do this, RRDtool will be able to store it. The values need to be
50 numerical but don't have to be, as opposed to MRTG, integers.
52 Many examples talk about SNMP which is an acronym for
53 Simple Network Management Protocol. "Simple" refers to the protocol --
54 it does not mean it is simple to manage or monitor a network. After working your
55 way through this document, you should know enough to be able to understand
56 what people are talking about. For now, just realize that SNMP is a way to
57 ask devices for the values of counters they keep.
58 It is the value from those counters that are kept in the RRD.
60 =head2 What can I do with this tool ?
62 RRDtool originated from MRTG (Multi Router Traffic Grapher). MRTG started
63 as a tiny little script for graphing the use of a connection
64 to the Internet. MRTG evolved into a tool for graphing other data sources
65 including temperature, speed, voltage, number of printouts and
66 the like. Most likely you will start to use the RRDtool to store
67 and process data collected via SNMP. The data will most likely be bytes
68 (or bits) transfered from and to a network or a computer.
69 RRDtool lets you create a database, store data in it, retrieve that data
70 and create graphs in PNG format for display on a web browser. Those PNG
71 images are dependent on the data you collected and could be, for instance,
72 an overview of the average network usage, or the peaks that occurred.
73 It can also be used to display tidal waves, solar radiation, power
74 consumption, number of visitors at an exhibition, noise levels near an
75 airport, temperature on your favorite holiday location, temperature in the
76 fridge and whatever you imagination can come up with. You need a sensor to
77 measure the data and be able to feed the numbers to RRDtool.
79 =head2 What if I still have problems after reading this document ?
81 First of all: read it again! You may have missed something.
82 If you are unable to compile the sources and you have a fairly common
83 OS, it will probably not be the fault of RRDtool. There may be pre-compiled
84 versions around on the Internet. If they come from trusted sources, get
86 If on the other hand the program works but does not give you the
87 expected results, it will be a problem with configuring it. Review
88 your configuration and compare it with the examples that follow.
90 There is a mailing list and an archive of it. Read the list for a few
91 weeks and search the archive. It is considered rude to just ask
92 a question without searching the archives: your problem may already have been
93 solved for somebody else! This is true for most, if not all, mailing lists
94 and not only for this particular list! Look in the documentation that
95 came with RRDtool for the location and usage of the list.
97 I suggest you take a moment to subscribe to the mailing list right now
98 by sending an email to E<lt>rrd-users-request@list.ee.ethz.chE<gt> with a
99 subject of "subscribe". If you ever want to leave this list, you write
100 an email to the same address but now with a subject of "unsubscribe".
102 =head2 How will you help me ?
104 By giving you some detailed descriptions with detailed examples.
105 It is assumed that following the instructions in the order presented
106 will give you enough knowledge of RRDtool to experiment for yourself.
107 If it doesn't work the first time, don't give up. Reread the stuff that
108 you did understand, you may have missed something.
109 By following the examples you get some hands-on experience and, even
110 more important, some background information of how it works.
112 You will need to know something about hexadecimal numbers. If you don't
113 then start with reading L<bin_dec_hex> before you continue here.
115 =head2 Your first Round Robin Database
117 In my opinion the best way to learn something is to actually do it.
118 Why not start right now? We will create a database, put some values
119 in it and extract this data again. Your output should be the same
120 as the output that is included in this document.
122 We will start with some easy stuff and compare a car with a router,
123 or compare kilometers (miles if you wish) with bits and bytes. It's
124 all the same: some number over some time.
126 Assume we have a device that transfers bytes to and from the Internet.
127 This device keeps a counter that starts at zero when it is turned on,
128 increasing with every byte that is transfered. This counter will have
129 a maximum value, if that value is reached and an extra byte is counted,
130 the counter starts all over at zero. This is the same as many counters
131 in the world such as the mileage counter in a car.
132 Most discussions about networking talk about bits per second so lets
133 get used to that right away. Assume a byte is eight bits and start to
134 think in bits not bytes. The counter, however, still counts bytes !
135 In the SNMP world most of the counters are 32 bits. That means they are
136 counting from 0 to 4294967295. We will use these values in the examples.
137 The device, when asked, returns the current value of the counter. We
138 know the time that has passes since we last asked so we now know how
139 many bytes have been transfered ***on average*** per second. This is
140 not very hard to calculate. First in words, then in calculations:
146 Take the current counter, subtract the previous value from it.
150 Do the same with the current time and the previous time.
154 Divide the outcome of (1) by the outcome of (2), the result is
155 the amount of bytes per second. Multiply by eight to get the
156 number of bits per second (bps).
160 bps = (counter_now - counter_before) / (time_now - time_before) * 8
162 For some people it may help to translate this to a automobile example:
163 Do not try this example, and if you do, don't blame me for the results.
165 People who are not used to think in kilometers per hour can translate
166 most into miles per hour by dividing km by 1.6 (close enough).
167 I will use the following abbreviations:
170 KM: kilometer (= 1000 meters).
173 KM/H: kilometers per hour
174 M/S: meters per second
176 You're driving a car. At 12:05 you read the counter in the dashboard
177 and it tells you that the car has moved 12345 KM until that moment.
178 At 12:10 you look again, it reads 12357 KM. This means you have
179 traveled 12 KM in five minutes. A scientist would translate that
180 into meters per second and this makes a nice comparison toward the
181 problem of (bytes per five minutes) versus (bits per second).
183 We traveled 12 kilometers which is 12000 meters. We did that in five
184 minutes which translates into 300 seconds. Our speed is 12000M / 300S
187 We could also calculate the speed in KM/H: 12 times five minutes
188 is an hour so we have to multiply 12 KM by 12 to get 144 KM/H.
189 For our native English speaking friends: that's 90 MPH so don't
190 try this example at home or where I live :)
192 Remember: these numbers are averages only. There is no way to figure out
193 from the numbers, if you drove at a constant speed. There is an example
194 later on in this tutorial that explains this.
196 I hope you understand that there is no difference in calculating M/S or
197 bps; only the way we collect the data is different. Even the K from kilo
198 is the same as in networking terms k also means 1000.
200 We will now create a database where we can keep all these interesting
201 numbers. The method used to start the program may differ slightly from
202 OS to OS but I assume you can figure it out if it works different on
203 your OS. Make sure you do not overwrite any file on your system when
204 executing the following command and type the whole line as one long
205 line (I had to split it for readability)
206 and skip all of the '\' characters.
208 rrdtool create test.rrd \
210 DS:speed:COUNTER:600:U:U \
211 RRA:AVERAGE:0.5:1:24 \
214 (So enter: C<rrdtool create test.rrd --start 920804400 DS ...>)
216 =head2 What has been created ?
218 We created the round robin database called test (test.rrd)
219 which starts at noon the day I started (7th of march, 1999) writing
220 this document. It holds one data source (DS) named "speed" that gets
221 built from a counter. This counter is read every five minutes (default)
222 In the same database two round robin archives (RRAs) are kept, one
223 averages the data every time it is read (e.g., there's nothing to average)
224 and keeps 24 samples (24 times 5 minutes is 2 hours). The other averages
225 6 values (half hour) and contains 10 of such averages (e.g., 5 hours)
226 The remaining options will be discussed later on.
228 RRDtool works with special time stamps coming from the UNIX world.
229 This time stamp is the number of seconds that passed since January
230 1st 1970 UTC. This time stamp is translated into local time and
231 it will therefore look different for the different time zones.
233 Chances are that you are not in the same part of the world as I am.
234 This means your time zone is different. In all examples where I talk
235 about time, the hours may be wrong for you. This has little effect on
236 the results of the examples, just correct the hours while reading.
237 As an example: where I will see "12:05" the UK folks will see "11:05".
239 We now have to fill our database with some numbers. We'll pretend to
240 have read the following numbers:
258 We fill the database as follows:
260 rrdtool update test.rrd 920804700:12345 920805000:12357 920805300:12363
261 rrdtool update test.rrd 920805600:12363 920805900:12363 920806200:12373
262 rrdtool update test.rrd 920806500:12383 920806800:12393 920807100:12399
263 rrdtool update test.rrd 920807400:12405 920807700:12411 920808000:12415
264 rrdtool update test.rrd 920808300:12420 920808600:12422 920808900:12423
266 This reads: update our test database with the following numbers
268 time 920804700, value 12345
269 time 920805000, value 12357
273 As you can see, it is possible to feed more than one value into the
274 database in one command. I had to stop at three for readability but
275 the real maximum is OS dependent.
277 We can now retrieve the data from our database using "rrdtool fetch":
279 rrdtool fetch test.rrd AVERAGE --start 920804400 --end 920809200
281 It should return the following output:
286 920805000: 4.0000000000e-02
287 920805300: 2.0000000000e-02
288 920805600: 0.0000000000e+00
289 920805900: 0.0000000000e+00
290 920806200: 3.3333333333e-02
291 920806500: 3.3333333333e-02
292 920806800: 3.3333333333e-02
293 920807100: 2.0000000000e-02
294 920807400: 2.0000000000e-02
295 920807700: 2.0000000000e-02
296 920808000: 1.3333333333e-02
297 920808300: 1.6666666667e-02
298 920808600: 6.6666666667e-03
299 920808900: 3.3333333333e-03
302 If it doesn't, something may be wrong. Perhaps your OS will print
303 "NaN" in a different form. It represents "Not A Number". If your OS
304 writes "U" or "UNKN" or something similar that's okay. If something
305 else is wrong, it will probably be due to an error you made (assuming
306 that my tutorial is correct of course :-). In that case: delete the
307 database and try again. Sometimes things change. This example used
308 to provide numbers like "0.04" in stead of "4.00000e-02". Those are
309 really the same numbers, just written down differently. Don't be
310 alarmed if a future version of rrdtool displays a slightly different
311 form of output. The examples in this document are correct for version
314 What this output represents will become clear in the rest of the tutorial.
316 =head2 It is time to create some graphics
318 Try the following command:
320 rrdtool graph speed.png \
321 --start 920804400 --end 920808000 \
322 DEF:myspeed=test.rrd:speed:AVERAGE \
325 This will create speed.png which starts at 12:00 and ends at 13:00.
326 There is a definition of variable myspeed, it is the data from RRA
327 "speed" out of database "test.rrd". The line drawn is 2 pixels high,
328 and comes from variable myspeed. The color is red.
329 You'll notice that the start of the graph is not at 12:00 but at 12:05
330 and this is because we have insufficient data to tell the average before
331 that time. This will only happen when you miss some samples, this will
332 not happen a lot, hopefully.
334 If this has worked: congratulations! If not, check what went wrong.
337 The colors are built up from red, green and blue. For each of the
338 components, you specify how much to use in hexadecimal where 00 means
339 not included and FF means fully included.
340 The "color" white is a mixture of red, green and blue: FFFFFF
341 The "color" black is all colors off: 000000
346 magenta #FF00FF (mixed red with blue)
347 gray #555555 (one third of all components)
349 Additionally you can add an alpha channel (transparency). The default
350 will be "FF" which means non-transparent.
352 The PNG you just created can be displayed using your favorite image
353 viewer. Web browsers will display the PNG via the URL
354 "file:///the/path/to/speed.png"
356 =head2 Graphics with some math
358 When looking at the image, you notice that the horizontal axis is labeled
359 12:10, 12:20, 12:30, 12:40 and 12:50. Sometimes a label doesn't fit (12:00
360 and 13:00 would be candidates) so they are skipped.
361 The vertical axis displays the range we entered. We provided kilometers
362 and when divided by 300 seconds, we get very small numbers. To be exact,
363 the first value was 12 (12357-12345) and divided by 300 this makes 0.04,
364 which is displayed by RRDtool as "40 m" meaning "40/1000". The "m" has
365 nothing to do with meters, kilometers or millimeters! RRDtool doesn't
366 know about all this, it just works with numbers and not with meters...
368 What we did wrong was that we should have measured in meters, this would
369 have been (12357000-12345000)/300 = 12000/300 = 40.
371 Let's correct that. We could recreate our database and store the correct
372 data but there is a better way: do some calculations while creating the
375 rrdtool graph speed2.png \
376 --start 920804400 --end 920808000 \
377 --vertical-label m/s \
378 DEF:myspeed=test.rrd:speed:AVERAGE \
379 CDEF:realspeed=myspeed,1000,\* \
380 LINE2:realspeed#FF0000
382 After viewing this PNG, you notice the "m" has disappeared. This it what
383 the correct result would be. Also, a label has been added to the image.
384 Apart from the things mentioned above, the PNG should be the same.
386 The calculations are in the CDEF part and are in Reverse Polish Notation
387 ("RPN"). What it says is: "take the data source myspeed and the number
388 1000; multiply those". Don't bother with RPN yet, it will be explained
389 later on in more detail. Also, you may want to read my tutorial on CDEFs
390 and Steve Rader's tutorial on RPN. But first finish this tutorial.
392 Hang on! If we can multiply values with 1000, it should also be possible
393 to display kilometers per hour from the same data!
395 To change a value that is measured in meters per second:
396 -*- Calculate meters per hour: value * 3600
397 -*- Calculate kilometers per hour: value / 1000
398 -*- Together this makes: value * (3600/1000) == value * 3.6
400 In our example database we made a mistake and we need to compensate for
401 this by multiplying with 1000. Applying that correction:
402 -*- value * 3.6 *1000 == value * 3600
404 Now let's create this PNG, and add some more magic ...
406 rrdtool graph speed3.png \
407 --start 920804400 --end 920808000 \
408 --vertical-label km/h \
409 DEF:myspeed=test.rrd:speed:AVERAGE \
410 "CDEF:kmh=myspeed,3600,*" \
411 CDEF:fast=kmh,100,GT,kmh,0,IF \
412 CDEF:good=kmh,100,GT,0,kmh,IF \
413 HRULE:100#0000FF:"Maximum allowed" \
414 AREA:good#00FF00:"Good speed" \
415 AREA:fast#FF0000:"Too fast"
417 This looks much better. Speed in KM/H and even an extra line with the
418 maximum allowed speed (on the road I travel at). I also changed the
419 colors used to display speed and changed it from a line into an area.
421 The calculations are more complex now. For the "good" speed they are:
423 Check if kmh is greater than 100 ( kmh,100 ) GT
424 If so, return 0, else kmh ((( kmh,100 ) GT ), 0, kmh) IF
428 Check if kmh is greater than 100 ( kmh,100 ) GT
429 If so, return kmh, else return 0 ((( kmh,100) GT ), kmh, 0) IF
431 =head2 Graphics Magic
433 I like to believe there are virtually no limits to how RRDtool graph
434 can manipulate data. I will not explain how it works, but look at the
437 rrdtool graph speed4.png \
438 --start 920804400 --end 920808000 \
439 --vertical-label km/h \
440 DEF:myspeed=test.rrd:speed:AVERAGE \
441 "CDEF:kmh=myspeed,3600,*" \
442 CDEF:fast=kmh,100,GT,100,0,IF \
443 CDEF:over=kmh,100,GT,kmh,100,-,0,IF \
444 CDEF:good=kmh,100,GT,0,kmh,IF \
445 HRULE:100#0000FF:"Maximum allowed" \
446 AREA:good#00FF00:"Good speed" \
447 AREA:fast#550000:"Too fast" \
448 STACK:over#FF0000:"Over speed"
450 Let's create a quick and dirty HTML page to view three PNGs:
452 <HTML><HEAD><TITLE>Speed</TITLE></HEAD><BODY>
453 <IMG src="speed2.png" alt="Speed in meters per second">
455 <IMG src="speed3.png" alt="Speed in kilometers per hour">
457 <IMG src="speed4.png" alt="Traveled too fast?">
460 Name the file "speed.html" or similar, and view it.
462 Now, all you have to do is measure the values regularly and update the
463 database. When you want to view the data, recreate the PNGs and make
464 sure to refresh them in your browser. (Note: just clicking reload may
465 not be enough, especially when proxies are involved. Try shift-reload
468 =head2 Updates in Reality
470 We've already used the "update" command: it took one or more parameters
471 in the form of "<time>:<value>". You'll be glad to know that you can
472 get the current time by filling in a "N" as the time.
473 If you wish, you can also use the "time" function in perl.
474 The shortest example in this doc :)
476 perl -e 'print time, "\n" '
478 How you can run a program on regular intervals is OS specific. But here's
479 an example in pseudo code:
481 Get the value, put it in variable "$speed"
482 rrdtool update speed.rrd N:$speed
484 (Do not try this with our test database, it is used in further examples)
486 This is all. Run this script every five minutes. When you need to know
487 what the graphics look like, run the examples above. You could put them
488 in a script. After running that script, view index.html
490 =head2 Some words on SNMP
492 I can imagine very few people will be able to get real data from their
493 car every five minutes, all other people will have to settle for some
494 other kind of counter. You could measure the number of pages printed by
495 a printer, the coffee made by the coffee machine, a device that counts
496 the electricity used, whatever. Any incrementing counter can be monitored
497 and graphed using the stuff you learned until now. Later on we will also
498 be able to monitor other types of values like temperature.
499 Most people will use the counter that keeps track
500 of octets (bytes) transfered by a network device so we have to do just
501 that. We will start with a description of how to collect data.
502 Some people will make a remark that there are tools which can do this data
503 collection for you. They are right! However, I feel it is important that
504 you understand they are not necessary. When you have to determine why
505 things went wrong you need to know how they work.
507 One tool used in the example has been talked about very briefly in the
508 beginning of this document, it is called SNMP. It is a way of talking to
509 equipment. The tool I use below is called "snmpget" and this is how it
512 snmpget device password OID
516 snmpget -v[version] -c[password] device OID
518 For device you substitute the name, or the IP address, of your device.
519 For password you use the "community read string" as it is called in the
520 SNMP world. For some devices the default of "public" might work, however
521 this can be disabled, altered or protected for privacy and security
522 reasons. Read the documentation that comes with your device or program.
524 Then there is this parameter, called OID, which means "object identifier".
526 When you start to learn about SNMP it looks very confusing. It isn't
527 all that difficult when you look at the Management Information Base
528 ("MIB"). It is an upside-down tree that describes data, with a single node
529 as the root and from there a number of branches. These branches end
530 up in another node, they branch out, etc. All the branches have a name
531 and they form the path that we follow all the way down. The branches
532 that we follow are named: iso, org, dod, internet, mgmt and mib-2.
533 These names can also be written down as numbers and are 1 3 6 1 2 1.
535 iso.org.dod.internet.mgmt.mib-2 (1.3.6.1.2.1)
537 There is a lot of confusion about the leading dot that some programs
538 use. There is *no* leading dot in an OID. However, some programs
539 can use above part of OIDs as a default. To indicate the difference
540 between abbreviated OIDs and full OIDs they need a leading dot when
541 you specify the complete OID. Often those programs will leave out
542 the default portion when returning the data to you. To make things
543 worse, they have several default prefixes ...
545 Right, lets continue to the start of our OID: we had 1.3.6.1.2.1
546 From there, we are especially interested in the branch "interfaces"
547 which has number 2 (e.g., 1.3.6.1.2.1.2 or 1.3.6.1.2.1.interfaces).
549 First, we have to get some SNMP program. First look if there is a
550 pre-compiled package available for your OS. This is the preferred way.
551 If not, you will have to get yourself the sources and compile those.
552 The Internet is full of sources, programs etc. Find information using
553 a search engine or whatever you prefer.
555 Assume you got the program. First try to collect some data that is
556 available on most systems. Remember: there is a short name for the
557 part of the tree that interests us most in the world we live in!
559 I will give an example which can be used on Fedora Core 3. If it
560 doesn't work for you, work your way through the manual of snmp and
561 adapt the example to make it work.
563 snmpget -v2c -c public myrouter system.sysDescr.0
565 The device should answer with a description of itself, perhaps empty.
566 Until you got a valid answer from a device, perhaps using a different
567 "password", or a different device, there is no point in continuing.
569 snmpget -v2c -c public myrouter interfaces.ifNumber.0
571 Hopefully you get a number as a result, the number of interfaces.
572 If so, you can carry on and try a different program called "snmpwalk".
574 snmpwalk -v2c -c public myrouter interfaces.ifTable.ifEntry.ifDescr
576 If it returns with a list of interfaces, you're almost there.
578 [user@host /home/alex]$ snmpwalk -v2c -c public cisco 2.2.1.2
580 interfaces.ifTable.ifEntry.ifDescr.1 = "BRI0: B-Channel 1"
581 interfaces.ifTable.ifEntry.ifDescr.2 = "BRI0: B-Channel 2"
582 interfaces.ifTable.ifEntry.ifDescr.3 = "BRI0" Hex: 42 52 49 30
583 interfaces.ifTable.ifEntry.ifDescr.4 = "Ethernet0"
584 interfaces.ifTable.ifEntry.ifDescr.5 = "Loopback0"
586 On this cisco equipment, I would like to monitor the "Ethernet0"
587 interface and see that it is number four. I try:
589 [user@host /home/alex]$ snmpget -v2c -c public cisco 2.2.1.10.4 2.2.1.16.4
591 interfaces.ifTable.ifEntry.ifInOctets.4 = 2290729126
592 interfaces.ifTable.ifEntry.ifOutOctets.4 = 1256486519
594 So now I have two OIDs to monitor and they are (in full, this time):
602 both with an interface number of 4.
604 Don't get fooled, this wasn't my first try. It took some time for me too
605 to understand what all these numbers mean, it does help a lot when they
606 get translated into descriptive text... At least, when people are talking
607 about MIBs and OIDs you know what it's all about.
608 Do not forget the interface number (0 if it is not interface dependent)
609 and try snmpwalk if you don't get an answer from snmpget.
611 If you understand above part, and get numbers from your device, continue
612 on with this tutorial. If not, then go back and re-read this part.
614 =head2 A Real World Example
616 Let the fun begin. First, create a new database. It contains data from
617 two counters, called input and output. The data is put into archives
618 that average it. They take 1, 6, 24 or 288 samples at a time.
619 They also go into archives that keep the maximum numbers. This will be
620 explained later on. The time in-between samples is 300 seconds, a good
621 starting point, which is the same as five minutes.
623 1 sample "averaged" stays 1 period of 5 minutes
624 6 samples averaged become one average on 30 minutes
625 24 samples averaged become one average on 2 hours
626 288 samples averaged become one average on 1 day
628 Lets try to be compatible with MRTG:
629 MRTG stores about the following amount of data:
631 600 5-minute samples: 2 days and 2 hours
632 600 30-minute samples: 12.5 days
633 600 2-hour samples: 50 days
634 732 1-day samples: 732 days
636 These ranges are appended so the total amount of data kept is approximately
637 797 days. RRDtool stores the data differently, it doesn't start the "weekly"
638 archive where the "daily" archive stopped. For both archives the most recent
639 data will be near "now" and therefore we will need to keep more data than
644 600 samples of 5 minutes (2 days and 2 hours)
645 700 samples of 30 minutes (2 days and 2 hours, plus 12.5 days)
646 775 samples of 2 hours (above + 50 days)
647 797 samples of 1 day (above + 732 days, rounded up to 797)
649 rrdtool create myrouter.rrd \
650 DS:input:COUNTER:600:U:U \
651 DS:output:COUNTER:600:U:U \
652 RRA:AVERAGE:0.5:1:600 \
653 RRA:AVERAGE:0.5:6:700 \
654 RRA:AVERAGE:0.5:24:775 \
655 RRA:AVERAGE:0.5:288:797 \
661 Next thing to do is collect data and store it. Here is an example.
662 It is written partially in pseudo code so you will have to find out what
663 to do exactly on your OS to make it work.
665 while not the end of the universe
668 snmpget router community 2.2.1.10.4
671 snmpget router community 2.2.1.16.4
674 rrdtool update myrouter.rrd N:$in:$out
679 Then, after collecting data for a day, try to create an image using:
681 rrdtool graph myrouter-day.png --start -86400 \
682 DEF:inoctets=myrouter.rrd:input:AVERAGE \
683 DEF:outoctets=myrouter.rrd:output:AVERAGE \
684 AREA:inoctets#00FF00:"In traffic" \
685 LINE1:outoctets#0000FF:"Out traffic"
687 This should produce a picture with one day worth of traffic.
688 One day is 24 hours of 60 minutes of 60 seconds: 24*60*60=86400, we
689 start at now minus 86400 seconds. We define (with DEFs) inoctets and
690 outoctets as the average values from the database myrouter.rrd and draw
691 an area for the "in" traffic and a line for the "out" traffic.
693 View the image and keep logging data for a few more days.
694 If you like, you could try the examples from the test database and
695 see if you can get various options and calculations working.
699 Display in bytes per second and in bits per second. Make the Ethernet
700 graphics go red if they are over four megabits per second.
702 =head2 Consolidation Functions
704 A few paragraphs back I mentioned the possibility of keeping
705 the maximum values instead of the average values. Let's go
706 into this a bit more.
709 Recall all the stuff about the speed of the car. Suppose we drove at 144
710 KM/H during 5 minutes and then were stopped by the police for 25 minutes.
711 At the end of the lecture we would take our laptop and create+view the
712 image taken from the database. If we look at the second RRA we did
713 create, we would have the average from 6 samples. The samples measured
714 would be 144+0+0+0+0+0=144, divided by 30 minutes, corrected for the
715 error by 1000, translated into KM/H, with a result of 24 KM/H.
716 I would still get a ticket but not for speeding anymore :)
718 Obviously, in this case, we shouldn't look at the averages. In some
719 cases they are handy. If you want to know how much KM you had traveled,
720 the picture would be the right one to look at. On the other hand, for
721 the speed that we traveled at, the maximum number seen is much more
722 valuable. (later we will see more types)
724 It is the same for data. If you want to know the amount, look at the
725 averages. If you want to know the rate, look at the maximum.
726 Over time, they will grow apart more and more. In the last database
727 we have created, there are two archives that keep data per day. The
728 archive that keeps averages will show low numbers, the archive that
729 shows maxima will have higher numbers.
730 For my car this would translate in averages per day of 96/24=4 KM/H
731 (as I travel about 94 kilometers on a day) during week days, and
732 maximum of 120 KM/H on weekdays (my top speed that I reach every day).
734 Big difference. Do not look at the second graph to estimate the
735 distances that I travel and do not look at the first graph to
736 estimate my speed. This will work if the samples are close together,
737 as they are in five minutes, but not if you average.
739 On some days, I go for a long ride. If I go across Europe and travel
740 for over 12 hours, the first graph will rise to about 60 KM/H. The
741 second one will show 180 KM/H. This means that I traveled a distance
742 of 60 KM/H times 24 H = 1440 KM. I did this with a higher speed and
743 a maximum around 180 KM/H. This doesn't mean that I traveled for 8
744 hours at a constant speed of 180 KM/H !
745 This is a real example: go with the flow through Germany (fast!) and stop
746 a few times for gas and coffee. Drive slowly through Austria and the
747 Netherlands. Be careful in the mountains and villages. If you would
748 look at the graphs created from the five-minute averages you would
749 get a totally different picture. You would see the same values on the
750 average and maximum graphs (provided I measured every 300 seconds).
751 You would be able to see when I stopped, when I was in top gear, when
752 I drove over fast highways etc. The granularity of the data is much
753 higher, so you can see more. However, this takes 12 samples per hour,
754 or 288 values per day, so it would be too much to keep for a long
755 period of time. Therefore we average it, eventually to one value per
756 day. From this one value, we cannot see much detail.
758 Make sure you understand the last few paragraphs. There is no value
759 in only a line and a few axis, you need to know what they mean and
760 interpret the data in a good way. This is true for all data.
762 The biggest mistake you can make is to use the collected data for
763 something that it is not suitable for. You would be better off if
764 you would not have the graphics at all in that case.
767 =head2 Let's review what you now should know.
769 You now know how to create a database. You can put the numbers in it,
770 get them out again by creating an image, do math on the data from the
771 database and view the outcome instead of the raw data.
772 You know about the difference between averages and maxima, and when
773 to use which (or at least you have an idea).
775 RRDtool can do more than what we have learned up to now. Before you
776 continue with the rest of this doc, I recommend that you reread from
777 the start and try some modifications on the examples. Make sure you
778 fully understand everything. It will be worth the effort and helps
779 you not only with the rest of this doc but also in your day to day
780 monitoring long after you read this introduction.
782 =head2 Data Source Types
784 All right, you feel like continuing. Welcome back and get ready
785 for an increased speed in the examples and explanation.
787 You know that in order to view a counter over time, you have to
788 take two numbers and divide the difference of them between the
789 time lapsed. This makes sense for the examples I gave you but there
790 are other possibilities. For instance, I'm able to retrieve the
791 temperature from my router in three places namely the inlet, the
792 so called hot-spot and the exhaust. These values are not counters.
793 If I take the difference of the two samples and divide that by
794 300 seconds I would be asking for the temperature change per second.
795 Hopefully this is zero! If not, the computer room is on fire :)
797 So, what can we do ? We can tell RRDtool to store the values we measure
798 directly as they are (this is not entirely true but close enough). The
799 graphs we make will look much better, they will show a rather constant
800 value. I know when the router is busy (it
801 works -> it uses more electricity -> it generates more heat -> the
802 temperature rises). I know when the doors are left open (the room is
803 cooled -> the warm air from the rest of the building flows into the
804 computer room -> the inlet temperature rises) etc. The data type we
805 use when creating the database before was counter, we now have a
806 different data type and thus a different name for it. It is called
807 GAUGE. There are more such data types:
809 - COUNTER we already know this one
810 - GAUGE we just learned this one
814 The two new types are DERIVE and ABSOLUTE. Absolute can be used like
815 counter with one difference: RRDtool assumes the counter is reset when
816 it's read. That is: its delta is known without calculation by RRDtool
817 whereas RRDtool needs to calculate it for the counter type.
818 Example: our first example (12345, 12357, 12363, 12363) would read:
819 unknown, 12, 6, 0. The rest of the calculations stay the same.
820 The other one, derive, is like counter. Unlike counter, it can also
821 decrease so it can have a negative delta. Again, the rest of the
822 calculations stay the same.
826 rrdtool create all.rrd --start 978300900 \
827 DS:a:COUNTER:600:U:U \
829 DS:c:DERIVE:600:U:U \
830 DS:d:ABSOLUTE:600:U:U \
832 rrdtool update all.rrd \
833 978301200:300:1:600:300 \
834 978301500:600:3:1200:600 \
835 978301800:900:5:1800:900 \
836 978302100:1200:3:2400:1200 \
837 978302400:1500:1:2400:1500 \
838 978302700:1800:2:1800:1800 \
839 978303000:2100:4:0:2100 \
840 978303300:2400:6:600:2400 \
841 978303600:2700:4:600:2700 \
842 978303900:3000:2:1200:3000
843 rrdtool graph all1.png -s 978300600 -e 978304200 -h 400 \
844 DEF:linea=all.rrd:a:AVERAGE LINE3:linea#FF0000:"Line A" \
845 DEF:lineb=all.rrd:b:AVERAGE LINE3:lineb#00FF00:"Line B" \
846 DEF:linec=all.rrd:c:AVERAGE LINE3:linec#0000FF:"Line C" \
847 DEF:lined=all.rrd:d:AVERAGE LINE3:lined#000000:"Line D"
849 =head2 RRDtool under the Microscope
855 Line A is a counter so it should continuously increment and RRDtool
856 should calculate the differences. Also, RRDtool needs to divide the
857 difference by the amount of time lapsed. This should end up as a
858 straight line at 1 (the deltas are 300, the time is 300).
862 Line B is of type GAUGE. These are "real" values so they should match
863 what we put in: a sort of a wave.
867 Line C is derive. It should be a counter that can decrease. It does
868 so between 2400 and 0, with 1800 in-between.
872 Line D is of type absolute. This is like counter but it works on
873 values without calculating the difference. The numbers are the same
874 and as you can see (hopefully) this has a different result.
878 This translates in the following values, starting at 23:10 and ending
879 at 00:10 the next day (where U means unknown/unplotted):
881 - Line A: u u 1 1 1 1 1 1 1 1 1 u
882 - Line B: u 1 3 5 3 1 2 4 6 4 2 u
883 - Line C: u u 2 2 2 0 -2 -6 2 0 2 u
884 - Line D: u 1 2 3 4 5 6 7 8 9 10 u
886 If your PNG shows all this, you know you have typed the data correct,
887 the RRDtool executable is working properly, your viewer doesn't fool you
888 and you successfully entered the year 2000 :)
889 You could try the same example four times, each time with only one of
892 Let's go over the data again:
898 Line A: 300,600,900 and so on. The counter delta is a constant 300 and
899 so it the time delta. A number divided by itself is always 1 (except
900 when dividing by zero which is undefined/illegal).
901 Why is it that the first point is unknown ? We do know what we put into
902 the database ? True ! But we didn't have a value to calculate the delta
903 from so we don't know where we started. It would be wrong to assume we
904 started at zero so we don't !
908 Line B: There is nothing to calculate. The numbers are as is.
912 Line C: Again, the start-out value is unknown. The same story is valid
913 like for line A. In this case the deltas are not constant so the line
914 is not. If we would put the same numbers in the database as we did for
915 line A, we would have gotten the same line. Unlike type counter,
916 this type can decrease and I hope to show you later on why
917 there is a difference.
921 Line D: Here the device calculates the deltas. Therefore we DO know the
922 first delta and it is plotted. We had the same input as with line A but
923 the meaning of this input is different. Therefore the line is different.
924 In this case the deltas increase each time with 300. The time delta
925 stays at a constant 300 and therefore the division of the two gives
932 There are a few more basics to show. Some important options are still to
933 be covered and we haven't look at counter wraps yet. First the counter wrap:
934 In our car we notice that our counter shows 999987. We travel 20 KM and
935 the counter should go to 1000007. Unfortunately, there are only six digits
936 on our counter so it really shows 000007. If we would plot that on a type
937 DERIVE, it would mean that the counter was set back 999980 KM. It wasn't,
938 and there has to be some protection for this. This protection is only
939 available for type COUNTER which should be used for this kind of counter
940 anyways. How does it work ? Type counter should never decrease and
941 therefore RRDtool must assume it wrapped if it does decrease !
942 If the delta is negative, this can be compensated for by adding the
943 maximum value of the counter + 1. For our car this would be:
945 Delta = 7 - 999987 = -999980 (instead of 1000007-999987=20)
947 Real delta = -999980 + 999999 + 1 = 20
949 At the time of writing this document, RRDtool knows of counters that
950 are either 32 bits or 64 bits of size. These counters can handle the
951 following different values:
953 - 32 bits: 0 .. 4294967295
954 - 64 bits: 0 .. 18446744073709551615
956 If these numbers look strange to you, you would like to view them in
957 their hexadecimal form:
959 - 32 bits: 0 .. FFFFFFFF
960 - 64 bits: 0 .. FFFFFFFFFFFFFFFF
962 RRDtool handles both counters the same. If an overflow occurs and
963 the delta would be negative, RRDtool first adds the maximum of a small
964 counter + 1 to the delta. If the delta is still negative, it had to be
965 the large counter that wrapped. Add the maximum possible value of the
966 large counter + 1 and subtract the falsely added small value.
967 There is a risk in this: suppose the large counter wrapped while adding
968 a huge delta, it could happen in theory that adding the smaller value
969 would make the delta positive. In this unlikely case the results would
970 not be correct. The increase should be nearly as high as the maximum
971 counter value for that to happen so chances are you would have several
972 other problems as well and this particular problem would not even be
973 worth thinking about. Even though I did include an example of it so you
974 can judge that for yourself.
976 The next section gives you some numerical examples for counter-wraps.
977 Try to do the calculations yourself or just believe me if your calculator
978 can't handle the numbers :)
982 - 32 bits: (4294967295+1) = 4294967296
983 - 64 bits: (18446744073709551615+1)-correction1 = 18446744069414584320
987 Should become: 4294967300
990 Correction1: -4294967196 +4294967296 = 100
992 Before: 18446744073709551000
994 Should become: 18446744073709551800
996 Delta: -18446744073709550816
997 Correction1: -18446744073709550816 +4294967296 = -18446744069414583520
998 Correction2: -18446744069414583520 +18446744069414584320 = 800
1000 Before: 18446744073709551615 ( maximum value )
1001 Increase: 18446744069414584320 ( absurd increase, minimum for
1002 Should become: 36893488143124135935 this example to work )
1003 But really is: 18446744069414584319
1005 Correction1: -4294967296 + 4294967296 = 0
1006 (not negative -> no correction2)
1008 Before: 18446744073709551615 ( maximum value )
1009 Increase: 18446744069414584319 ( one less increase )
1010 Should become: 36893488143124135934
1011 But really is: 18446744069414584318
1013 Correction1: -4294967297 +4294967296 = -1
1014 Correction2: -1 +18446744069414584320 = 18446744069414584319
1016 As you can see from the last two examples, you need strange numbers
1017 for RRDtool to fail (provided it's bug free of course) so this should
1018 not happen. However, SNMP or whatever method you choose to collect the
1019 data might also report wrong numbers occasionally. We can't prevent all
1020 errors but there are some things we can do. The RRDtool "create" command
1021 takes two special parameters for this. They define
1022 the minimum and maximum allowed value. Until now, we used "U", meaning
1023 "unknown". If you provide values for one or both of them and if RRDtool
1024 receives values that are outside these limits, it will ignore those
1025 values. For a thermometer in degrees Celsius, the absolute minimum is
1026 just under -273. For my router, I can assume this minimum is much higher
1027 so I would say it is 10. The maximum temperature for my router I would
1028 state as 80. Any higher and the device would be out of order.
1029 For my car, I would never expect negative numbers and also I would not
1030 expect numbers to be higher than 230. Anything else, and there must have
1031 been an error. Remember: the opposite is not true, if the numbers pass
1032 this check it doesn't mean that they are correct. Always judge the
1033 graph with a healthy dose of paranoia if it looks weird.
1035 =head2 Data Resampling
1037 One important feature of RRDtool has not been explained yet:
1038 It is virtually impossible to collect the data and feed it into RRDtool
1039 on exact intervals. RRDtool therefore interpolates the data so it is on
1040 exact intervals. If you do not know what this means or how it works,
1041 then here's the help you seek:
1043 Suppose a counter increases with exactly one for every second. You want
1044 to measure it in 300 seconds intervals. You should retrieve values
1045 that are exactly 300 apart. However, due to various circumstances you
1046 are a few seconds late and the interval is 303. The delta will also be
1047 303 in that case. Obviously RRDtool should not put 303 in the database
1048 and make you believe that the counter increased 303 in 300 seconds.
1049 This is where RRDtool interpolates: it alters the 303 value as if it
1050 would have been stored earlier and it will be 300 in 300 seconds.
1051 Next time you are at exactly the right time. This means that the current
1052 interval is 297 seconds and also the counter increased with 297. Again
1053 RRDtool alters the value and stores 300 as it should be.
1055 in the RRD in reality
1057 time+000: 0 delta="U" time+000: 0 delta="U"
1058 time+300: 300 delta=300 time+300: 300 delta=300
1059 time+600: 600 delta=300 time+603: 603 delta=303
1060 time+900: 900 delta=300 time+900: 900 delta=297
1062 Let's create two identical databases. I've chosen the time range 920805000
1063 to 920805900 as this goes very well with the example numbers.
1065 rrdtool create seconds1.rrd \
1067 DS:seconds:COUNTER:600:U:U \
1068 RRA:AVERAGE:0.5:1:24
1072 for Unix: cp seconds1.rrd seconds2.rrd
1073 for Dos: copy seconds1.rrd seconds2.rrd
1074 for vms: how would I know :)
1078 rrdtool update seconds1.rrd \
1079 920805000:000 920805300:300 920805600:600 920805900:900
1080 rrdtool update seconds2.rrd \
1081 920805000:000 920805300:300 920805603:603 920805900:900
1085 rrdtool graph seconds1.png \
1086 --start 920804700 --end 920806200 \
1088 --upper-limit 1.05 --lower-limit 0.95 --rigid \
1089 DEF:seconds=seconds1.rrd:seconds:AVERAGE \
1090 CDEF:unknown=seconds,UN \
1091 LINE2:seconds#0000FF \
1093 rrdtool graph seconds2.png \
1094 --start 920804700 --end 920806200 \
1096 --upper-limit 1.05 --lower-limit 0.95 --rigid \
1097 DEF:seconds=seconds2.rrd:seconds:AVERAGE \
1098 CDEF:unknown=seconds,UN \
1099 LINE2:seconds#0000FF \
1102 View both images together (add them to your index.html file)
1103 and compare. Both graphs should show the same, despite the
1104 input being different.
1108 It's time to wrap up this document. You now know all the basics to be
1109 able to work with RRDtool and to read the documentation available.
1110 There is plenty more to discover about RRDtool and you will find more and
1111 more uses for the package. You could create easy graphics using just the
1112 examples provided and using only RRDtool. You could also use the front
1113 ends that are available.
1117 Remember to subscribe to the mailing-list. Even if you are not answering
1118 the mails that come by, it helps both you and the rest. A lot of the stuff
1119 that I know about MRTG (and therefore about RRDtool) I've learned while
1120 just reading the list without posting to it. I did not need to ask the
1121 basic questions as they are answered in the FAQ (read it!) and
1122 in various mails by other users.
1123 With thousands of users all over the world, there will always be people
1124 who ask questions that you can answer because you read this and other
1125 documentation and they didn't.
1129 The RRDtool manpages
1133 I hope you enjoyed the examples and their descriptions. If you do, help
1134 other people by pointing them to this document when they are asking
1135 basic questions. They will not only get their answer but at the same
1136 time learn a whole lot more.
1138 Alex van den Bogaerdt
1139 E<lt>alex@ergens.op.het.netE<gt>
projects / rrdtool.git / blob