Code

Imported upstream SVN snapshot 1.4~rc2+20090928.
[pkg-rrdtool.git] / doc / rrdcached.html
1 <?xml version="1.0" ?>
2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
3 <html xmlns="http://www.w3.org/1999/xhtml">
4 <head>
5 <title>rrdcached</title>
6 <meta http-equiv="content-type" content="text/html; charset=utf-8" />
7 <link rev="made" href="mailto:root@localhost" />
8 </head>
10 <body style="background-color: white">
13 <!-- INDEX BEGIN -->
14 <div name="index">
15 <p><a name="__index__"></a></p>
16 <!--
18 <ul>
20         <li><a href="#name">NAME</a></li>
21         <li><a href="#synopsis">SYNOPSIS</a></li>
22         <li><a href="#description">DESCRIPTION</a></li>
23         <li><a href="#options">OPTIONS</a></li>
24         <li><a href="#affected_rrdtool_commands">AFFECTED RRDTOOL COMMANDS</a></li>
25         <li><a href="#error_reporting">ERROR REPORTING</a></li>
26         <li><a href="#how_it_works">HOW IT WORKS</a></li>
27         <li><a href="#security_considerations">SECURITY CONSIDERATIONS</a></li>
28         <ul>
30                 <li><a href="#authentication">Authentication</a></li>
31                 <li><a href="#authorization">Authorization</a></li>
32                 <li><a href="#encryption">Encryption</a></li>
33                 <li><a href="#sanity_checking">Sanity checking</a></li>
34                 <li><a href="#conclusion">Conclusion</a></li>
35         </ul>
37         <li><a href="#protocol">PROTOCOL</a></li>
38         <ul>
40                 <li><a href="#valid_commands">Valid Commands</a></li>
41                 <li><a href="#performance_values">Performance Values</a></li>
42         </ul>
44         <li><a href="#signals">SIGNALS</a></li>
45         <li><a href="#bugs">BUGS</a></li>
46         <li><a href="#see_also">SEE ALSO</a></li>
47         <li><a href="#author">AUTHOR</a></li>
48         <li><a href="#contributors">CONTRIBUTORS</a></li>
49 </ul>
51 -->
54 </div>
55 <!-- INDEX END -->
57 <p>
58 </p>
59 <hr />
60 <h1><a name="name">NAME</a></h1>
61 <p>rrdcached - Data caching daemon for rrdtool</p>
62 <p>
63 </p>
64 <hr />
65 <h1><a name="synopsis">SYNOPSIS</a></h1>
66 <p><strong>rrdcached</strong>
67 [<strong>-P</strong>&nbsp;<em>permissions</em>]
68 [<strong>-l</strong>&nbsp;<em>address</em>]
69 [<strong>-w</strong>&nbsp;<em>timeout</em>]
70 [<strong>-z</strong>&nbsp;<em>delay</em>]
71 [<strong>-f</strong>&nbsp;<em>timeout</em>]
72 [<strong>-p</strong>&nbsp;<em>pid_file</em>]
73 [<strong>-t</strong>&nbsp;<em>write_threads</em>]
74 [<strong>-j</strong>&nbsp;<em>journal_dir</em>]
75 [-F]
76 [-g]
77 [<strong>-b</strong>&nbsp;<em>base_dir</em>&nbsp;[<strong>-B</strong>]]</p>
78 <p>
79 </p>
80 <hr />
81 <h1><a name="description">DESCRIPTION</a></h1>
82 <p><strong>rrdcached</strong> is a daemon that receives updates to existing RRD files,
83 accumulates them and, if enough have been received or a defined time has
84 passed, writes the updates to the RRD file. A <em>flush</em> command may be used to
85 force writing of values to disk, so that graphing facilities and similar can
86 work with up-to-date data.</p>
87 <p>The daemon was written with big setups in mind. Those setups usually run into
88 IO&nbsp;related problems sooner or later for reasons that are beyond the scope
89 of this document. Check the wiki at the RRDTool homepage for details. Also
90 check <a href="#security_considerations">SECURITY CONSIDERATIONS</a> below before using this daemon! A detailed
91 description of how the daemon operates can be found in the <a href="#how_it_works">HOW IT WORKS</a>
92 section below.</p>
93 <p>
94 </p>
95 <hr />
96 <h1><a name="options">OPTIONS</a></h1>
97 <dl>
98 <dt><strong><a name="l_address" class="item"><strong>-l</strong> <em>address</em></a></strong></dt>
100 <dd>
101 <p>Tells the daemon to bind to <em>address</em> and accept incoming connections on that
102 socket. If <em>address</em> begins with <code>unix:</code>, everything following that prefix is
103 interpreted as the path to a UNIX domain socket. Otherwise the address or node
104 name are resolved using <em>getaddrinfo</em>.</p>
105 <p>For network sockets, a port may be specified by using the form
106 <code>[address]:port</code>. If the address is an IPv4 address or a fully
107 qualified domain name (i.&nbsp;e. the address contains at least one dot
108 (<code>.</code>)), the square brackets can be omitted, resulting in the (simpler)
109 <code>address:port</code> pattern. The default port is <strong>42217/udp</strong>. If you
110 specify a network socket, it is mandatory to read the
111 <a href="#security_considerations">SECURITY CONSIDERATIONS</a> section.</p>
112 <p>The following formats are accepted. Please note that the address of the UNIX
113 domain socket <strong>must</strong> start with a slash in the second case!</p>
114 <pre>
115    unix:&lt;/path/to/unix.sock&gt;
116    /&lt;path/to/unix.sock&gt;
117    &lt;hostname-or-ip&gt;
118    [&lt;hostname-or-ip&gt;]:&lt;port&gt;
119    &lt;hostname-or-ipv4&gt;:&lt;port&gt;</pre>
120 <p>If the <strong>-l</strong> option is not specified the default address,
121 <code>unix:/tmp/rrdcached.sock</code>, will be used.</p>
122 </dd>
123 <dt><strong><a name="p_command_command" class="item"><strong>-P</strong> <em>command</em>[,<em>command</em>[,...]]</a></strong></dt>
125 <dd>
126 <p>Specifies the commands accepted via a network socket. This allows
127 administrators of <em>RRDCacheD</em> to control the actions accepted from various
128 sources.</p>
129 <p>The arguments given to the <strong>-P</strong> option is a comma separated list of commands.
130 For example, to allow the <code>FLUSH</code> and <code>PENDING</code> commands one could specify:</p>
131 <pre>
132   rrdcached -P FLUSH,PENDING $MORE_ARGUMENTS</pre>
133 <p>The <strong>-P</strong> option effects the <em>following</em> socket addresses (the following <strong>-l</strong>
134 options). In the following example, only the IPv4 network socket (address
135 <code>10.0.0.1</code>) will be restricted to the <code>FLUSH</code> and <code>PENDING</code> commands:</p>
136 <pre>
137   rrdcached -l unix:/some/path -P FLUSH,PENDING -l 10.0.0.1</pre>
138 <p>A complete list of available commands can be found in the section
139 <a href="#valid_commands">Valid Commands</a> below. There are two minor special exceptions:</p>
140 <ul>
141 <li>
142 <p>The <code>HELP</code> and <a href="#quit"><code>QUIT</code></a> commands are always allowed.</p>
143 </li>
144 <li>
145 <p>If the <a href="#batch"><code>BATCH</code></a> command is accepted, the <strong>.</strong>&nbsp;command will automatically
146 be accepted, too.</p>
147 </li>
148 </ul>
149 <p>Please also read <a href="#security_considerations">SECURITY CONSIDERATIONS</a> below.</p>
150 </dd>
151 <dt><strong><a name="w_timeout" class="item"><strong>-w</strong> <em>timeout</em></a></strong></dt>
153 <dd>
154 <p>Data is written to disk every <em>timeout</em> seconds. If this option is not
155 specified the default interval of 300&nbsp;seconds will be used.</p>
156 </dd>
157 <dt><strong><a name="z_delay" class="item"><strong>-z</strong> <em>delay</em></a></strong></dt>
159 <dd>
160 <p>If specified, rrdcached will delay writing of each RRD for a random number
161 of seconds in the range&nbsp;[0,<em>delay</em>).  This will avoid too many
162 writes being queued simultaneously.  This value should be no greater than
163 the value specified in <strong>-w</strong>.  By default, there is no delay.</p>
164 </dd>
165 <dt><strong><a name="f_timeout" class="item"><strong>-f</strong> <em>timeout</em></a></strong></dt>
167 <dd>
168 <p>Every <em>timeout</em> seconds the entire cache is searched for old values which are
169 written to disk. This only concerns files to which updates have stopped, so
170 setting this to a high value, such as 3600&nbsp;seconds, is acceptable in most
171 cases. This timeout defaults to 3600&nbsp;seconds.</p>
172 </dd>
173 <dt><strong><a name="p_file" class="item"><strong>-p</strong> <em>file</em></a></strong></dt>
175 <dd>
176 <p>Sets the name and location of the PID-file. If not specified, the default,
177 <code>$localststedir/run/rrdcached.pid</code> will be used.</p>
178 </dd>
179 <dt><strong><a name="t_write_threads" class="item"><strong>-t</strong> <em>write_threads</em></a></strong></dt>
181 <dd>
182 <p>Specifies the number of threads used for writing RRD files.  The default
183 is&nbsp;4.  Increasing this number will allow rrdcached to have more
184 simultaneous I/O requests into the kernel.  This may allow the kernel to
185 re-order disk writes, resulting in better disk throughput.</p>
186 </dd>
187 <dt><strong><a name="j_dir" class="item"><strong>-j</strong> <em>dir</em></a></strong></dt>
189 <dd>
190 <p>Write updates to a journal in <em>dir</em>.  In the event of a program or system
191 crash, this will allow the daemon to write any updates that were pending
192 at the time of the crash.</p>
193 <p>On startup, the daemon will check for journal files in this directory.  If
194 found, all updates therein will be read into memory before the daemon
195 starts accepting new connections.</p>
196 <p>The journal will be rotated with the same frequency as the flush timer
197 given by <strong>-f</strong>.</p>
198 <p>When journaling is enabled, the daemon will use a fast shutdown procedure.
199 Rather than flushing all files to disk, it will make sure the journal is
200 properly written and exit immediately.  Although the RRD data files are
201 not fully up-to-date, no information is lost; all pending updates will be
202 replayed from the journal next time the daemon starts up.</p>
203 <p>To disable fast shutdown, use the <strong>-F</strong> option.</p>
204 </dd>
205 <dt><strong><a name="f" class="item"><strong>-F</strong></a></strong></dt>
207 <dd>
208 <p>ALWAYS flush all updates to the RRD data files when the daemon is shut
209 down, regardless of journal setting.</p>
210 </dd>
211 <dt><strong><a name="g" class="item"><strong>-g</strong></a></strong></dt>
213 <dd>
214 <p>Run in the foreground.  The daemon will not <code>fork()</code>.</p>
215 </dd>
216 <dt><strong><a name="b_dir" class="item"><strong>-b</strong> <em>dir</em></a></strong></dt>
218 <dd>
219 <p>The daemon will change into a specific directory at startup. All files passed
220 to the daemon, that are specified by a <strong>relative</strong> path, will be interpreted
221 to be relative to this directory. If not given the default, <code>/tmp</code>, will be
222 used.</p>
223 <pre>
224   +------------------------+------------------------+
225   ! Command line           ! File updated           !
226   +------------------------+------------------------+
227   ! foo.rrd                ! /tmp/foo.rrd           !
228   ! foo/bar.rrd            ! /tmp/foo/bar.rrd       !
229   ! /var/lib/rrd/foo.rrd   ! /var/lib/rrd/foo.rrd   !
230   +------------------------+------------------------+
231   Paths given on the command  line and paths actually
232   updated by the daemon,  assuming the base directory
233   &quot;/tmp&quot;.</pre>
234 <p><strong>WARNING:</strong> The paths up to and including the base directory <strong>MUST NOT BE</strong>
235 symbolic links.  In other words, if the base directory is
236 specified as:</p>
237 <pre>
238     -b /base/dir/somewhere</pre>
239 <p>... then <strong>NONE</strong> of the following should be symbolic links:</p>
240 <pre>
241     /base
242     /base/dir
243     /base/dir/somewhere</pre>
244 </dd>
245 <dt><strong><a name="b" class="item"><strong>-B</strong></a></strong></dt>
247 <dd>
248 <p>Only permit writes into the base directory specified in <strong>-b</strong> (and any
249 sub-directories).  This does <strong>NOT</strong> detect symbolic links.  Paths
250 containing <code>../</code> will also be blocked.</p>
251 </dd>
252 </dl>
253 <p>
254 </p>
255 <hr />
256 <h1><a name="affected_rrdtool_commands">AFFECTED RRDTOOL COMMANDS</a></h1>
257 <p>The following commands may be made aware of the <strong>rrdcached</strong> using the command
258 line argument <strong>--daemon</strong> or the environment variable <strong>RRDCACHED_ADDRESS</strong>:</p>
259 <dl>
260 <dt><strong><a name="dump" class="item"><strong>dump</strong></a></strong></dt>
262 <dt><strong><a name="fetch" class="item"><strong>fetch</strong></a></strong></dt>
264 <dt><strong><a name="flush" class="item"><strong>flush</strong></a></strong></dt>
266 <dt><strong><a name="graph" class="item"><strong>graph</strong></a></strong></dt>
268 <dt><strong><a name="graphv" class="item"><strong>graphv</strong></a></strong></dt>
270 <dt><strong><a name="info" class="item"><strong>info</strong></a></strong></dt>
272 <dt><strong><a name="last" class="item"><strong>last</strong></a></strong></dt>
274 <dt><strong><a name="lastupdate" class="item"><strong>lastupdate</strong></a></strong></dt>
276 <dt><strong><a name="update" class="item"><strong>update</strong></a></strong></dt>
278 <dt><strong><a name="xport" class="item"><strong>xport</strong></a></strong></dt>
280 </dl>
281 <p>The <strong>update</strong> command can send values to the daemon instead of writing them to
282 the disk itself. All other commands can send a <strong>FLUSH</strong> command (see below) to
283 the daemon before accessing the files, so they work with up-to-date data even
284 if the cache timeout is large.</p>
285 <p>
286 </p>
287 <hr />
288 <h1><a name="error_reporting">ERROR REPORTING</a></h1>
289 <p>The daemon reports errors in one of two ways: During startup, error messages
290 are printed to <code>STDERR</code>. One of the steps when starting up is to fork to the
291 background and closing <code>STDERR</code> - after this writing directly to the user is
292 no longer possible. Once this has happened, the daemon will send log messages
293 to the system logging daemon using <em>syslog(3)</em>. The facility used is
294 <code>LOG_DAEMON</code>.</p>
295 <p>
296 </p>
297 <hr />
298 <h1><a name="how_it_works">HOW IT WORKS</a></h1>
299 <p>When receiving an update, <strong>rrdcached</strong> does not write to disk but looks for an
300 entry for that file in its internal tree. If not found, an entry is created
301 including the current time (called &quot;First&quot; in the diagram below). This time is
302 <strong>not</strong> the time specified on the command line but the time the operating system
303 considers to be &quot;now&quot;. The value and time of the value (called &quot;Time&quot; in the
304 diagram below) are appended to the tree node.</p>
305 <p>When appending a value to a tree node, it is checked whether it's time to write
306 the values to disk. Values are written to disk if
307 <code>now()&nbsp;-&nbsp;First&nbsp;&gt;=&nbsp;timeout</code>, where <code>timeout</code> is the timeout specified
308 using the <strong>-w</strong> option, see <em>OPTIONS</em>. If the values are &quot;old enough&quot; they
309 will be enqueued in the &quot;update queue&quot;, i.&nbsp;e. they will be appended to
310 the linked list shown below.  Because the tree nodes and the elements of the
311 linked list are the same data structures in memory, any update to a file that
312 has already been enqueued will be written with the next write to the RRD file,
313 too.</p>
314 <p>A separate &quot;update thread&quot; constantly dequeues the first element in the update
315 queue and writes all its values to the appropriate file. So as long as the
316 update queue is not empty files are written at the highest possible rate.</p>
317 <p>Since the timeout of files is checked only when new values are added to the
318 file, &quot;dead&quot; files, i.&nbsp;e. files that are not updated anymore, would never
319 be written to disk. Therefore, every now and then, controlled by the <strong>-f</strong>
320 option, the entire tree is walked and all &quot;old&quot; values are enqueued. Since this
321 only affects &quot;dead&quot; files and walking the tree is relatively expensive, you
322 should set the &quot;flush interval&quot; to a reasonably high value. The default is
323 3600&nbsp;seconds (one hour).</p>
324 <p>The downside of caching values is that they won't show up in graphs generated
325 from the RRD&nbsp;files. To get around this, the daemon provides the &quot;flush
326 command&quot; to flush specific files. This means that the file is inserted at the
327 <strong>head</strong> of the update queue or moved there if it is already enqueued. The flush
328 command will return only after the file's pending updates have been written
329 to disk.</p>
330 <pre>
331  +------+   +------+                               +------+
332  ! head !   ! root !                               ! tail !
333  +---+--+   +---+--+                               +---+--+
334      !         /\                                      !
335      !        /  \                                     !
336      !       /\  /\                                    !
337      !      /\/\ \ `----------------- ... --------,    !
338      V     /      `-------,                       !    V
339  +---+----+---+    +------+-----+             +---+----+---+
340  ! File:  foo !    ! File:  bar !             ! File:  qux !
341  ! First: 101 !    ! First: 119 !             ! First: 180 !
342  ! Next:&amp;bar -+---&gt;! Next:&amp;... -+---&gt; ... ---&gt;! Next:NULL  !
343  | Prev:NULL  !&lt;---+-Prev:&amp;foo  !&lt;--- ... ----+-Prev: &amp;... !
344  +============+    +============+             +============+
345  ! Time:  100 !    ! Time:  120 !             ! Time:  180 !
346  ! Value:  10 !    ! Value: 0.1 !             ! Value: 2,2 !
347  +------------+    +------------+             +------------+
348  ! Time:  110 !    ! Time:  130 !             ! Time:  190 !
349  ! Value:  26 !    ! Value: 0.1 !             ! Value: 7,3 !
350  +------------+    +------------+             +------------+
351  :            :    :            :             :            :
352  +------------+    +------------+             +------------+
353  ! Time:  230 !    ! Time:  250 !             ! Time:  310 !
354  ! Value:  42 !    ! Value: 0.2 !             ! Value: 1,2 !
355  +------------+    +------------+             +------------+</pre>
356 <p>The above diagram demonstrates:</p>
357 <ul>
358 <li>
359 <p>Files/values are stored in a (balanced) tree.</p>
360 </li>
361 <li>
362 <p>Tree nodes and entries in the update queue are the same data structure.</p>
363 </li>
364 <li>
365 <p>The local time (&quot;First&quot;) and the time specified in updates (&quot;Time&quot;) may differ.</p>
366 </li>
367 <li>
368 <p>Timed out values are inserted at the &quot;tail&quot;.</p>
369 </li>
370 <li>
371 <p>Explicitly flushed values are inserted at the &quot;head&quot;.</p>
372 </li>
373 <li>
374 <p>ASCII art rocks.</p>
375 </li>
376 </ul>
377 <p>
378 </p>
379 <hr />
380 <h1><a name="security_considerations">SECURITY CONSIDERATIONS</a></h1>
381 <p>
382 </p>
383 <h2><a name="authentication">Authentication</a></h2>
384 <p>There is no authentication.</p>
385 <p>The client/server protocol does not yet have any authentication mechanism. It
386 is likely that authentication and encryption will be added in a future version,
387 but for the time being it is the administrator's responsibility to secure the
388 traffic from/to the daemon!</p>
389 <p>It is highly recommended to install a packet filter or similar mechanism to
390 prevent unauthorized connections. Unless you have a dedicated VLAN or VPN for
391 this, using network sockets is probably a bad idea!</p>
392 <p>
393 </p>
394 <h2><a name="authorization">Authorization</a></h2>
395 <p>There is minimal per-socket authorization.</p>
396 <p>Authorization is currently done on a per-socket basis. That means each socket
397 has a list of commands it will accept and it will accept. It will accept only
398 those commands explicitly listed but it will (currently) accept these commands
399 from anyone reaching the socket.</p>
400 <p>If the networking sockets are to be used, it is necessary to restrict the
401 accepted commands to those needed by external clients. If, for example,
402 external clients want to draw graphs of the cached data, they should only be
403 allowed to use the <code>FLUSH</code> command.</p>
404 <p>
405 </p>
406 <h2><a name="encryption">Encryption</a></h2>
407 <p>There is no encryption.</p>
408 <p>Again, this may be added in the future, but for the time being it is your job
409 to keep your private data private. Install a VPN or an encrypted tunnel if you
410 statistics are confidential!</p>
411 <p>
412 </p>
413 <h2><a name="sanity_checking">Sanity checking</a></h2>
414 <p>There is no sanity checking.</p>
415 <p>The daemon will blindly write to any file it gets told, so you really should
416 create a separate user just for this daemon. Also it does not do any sanity
417 checks, so if it gets told to write values for a time far in the future, your
418 files will be messed up good!</p>
419 <p>
420 </p>
421 <h2><a name="conclusion">Conclusion</a></h2>
422 <ul>
423 <li>
424 <p>Security is the job of the administrator.</p>
425 </li>
426 <li>
427 <p>We recommend to allow write access via UNIX domain sockets only.</p>
428 </li>
429 <li>
430 <p>You have been warned.</p>
431 </li>
432 </ul>
433 <p>
434 </p>
435 <hr />
436 <h1><a name="protocol">PROTOCOL</a></h1>
437 <p>The daemon communicates with clients using a line based ASCII protocol which is
438 easy to read and easy to type. This makes it easy for scripts to implement the
439 protocol and possible for users to use <em>telnet</em> to connect to the daemon
440 and test stuff &quot;by hand&quot;.</p>
441 <p>The protocol is line based, this means that each record consists of one or more
442 lines. A line is terminated by the line feed character <code>0x0A</code>, commonly
443 written as <code>\n</code>. In the examples below, this character will be written as
444 <code>&lt;LF&gt;</code> (&quot;line feed&quot;).</p>
445 <p>After the connection has been established, the client is expected to send a
446 &quot;command&quot;. A command consists of the command keyword, possibly some arguments,
447 and a terminating newline character. For a list of commands, see
448 <a href="#valid_commands">Valid Commands</a> below.</p>
449 <p>Example:</p>
450 <pre>
451   FLUSH /tmp/foo.rrd&lt;LF&gt;</pre>
452 <p>The daemon answers with a line consisting of a status code and a short status
453 message, separated by one or more space characters. A negative status code
454 signals an error, a positive status code or zero signal success. If the status
455 code is greater than zero, it indicates the number of lines that follow the
456 status line.</p>
457 <p>Examples:</p>
458 <pre>
459  0 Success&lt;LF&gt;</pre>
460 <pre>
461  2 Two lines follow&lt;LF&gt;
462  This is the first line&lt;LF&gt;
463  And this is the second line&lt;LF&gt;</pre>
464 <p>
465 </p>
466 <h2><a name="valid_commands">Valid Commands</a></h2>
467 <p>The following commands are understood by the daemon:</p>
468 <dl>
469 <dt><strong><a name="flush_filename" class="item"><strong>FLUSH</strong> <em>filename</em></a></strong></dt>
471 <dd>
472 <p>Causes the daemon to put <em>filename</em> to the <strong>head</strong> of the update queue
473 (possibly moving it there if the node is already enqueued). The answer will be
474 sent <strong>after</strong> the node has been dequeued.</p>
475 </dd>
476 <dt><strong><a name="flushall" class="item"><strong>FLUSHALL</strong></a></strong></dt>
478 <dd>
479 <p>Causes the daemon to start flushing ALL pending values to disk.  This
480 returns immediately, even though the writes may take a long time.</p>
481 </dd>
482 <dt><strong><a name="pending_filename" class="item"><strong>PENDING</strong> <em>filename</em></a></strong></dt>
484 <dd>
485 <p>Shows any &quot;pending&quot; updates for a file, in order.  The updates shown have
486 not yet been written to the underlying RRD file.</p>
487 </dd>
488 <dt><strong><a name="forget_filename" class="item"><strong>FORGET</strong> <em>filename</em></a></strong></dt>
490 <dd>
491 <p>Removes <em>filename</em> from the cache.  Any pending updates <strong>WILL BE LOST</strong>.</p>
492 </dd>
493 <dt><strong><a name="queue" class="item"><strong>QUEUE</strong></a></strong></dt>
495 <dd>
496 <p>Shows the files that are on the output queue.  Returns zero or more lines
497 in the following format, where &lt;num_vals&gt; is the number of values
498 to be written for the &lt;file&gt;:</p>
499 <pre>
500     &lt;num_vals&gt; &lt;file&gt;</pre>
501 </dd>
502 <dt><strong><a name="help_command" class="item"><strong>HELP</strong> [<em>command</em>]</a></strong></dt>
504 <dd>
505 <p>Returns a short usage message. If no command is given, or <em>command</em> is
506 <strong>HELP</strong>, a list of commands supported by the daemon is returned. Otherwise a
507 short description, possibly containing a pointer to a manual page, is returned.
508 Obviously, this is meant for interactive usage and the format in which the
509 commands and usage summaries are returned is not well defined.</p>
510 </dd>
511 <dt><strong><a name="stats" class="item"><strong>STATS</strong></a></strong></dt>
513 <dd>
514 <p>Returns a list of metrics which can be used to measure the daemons performance
515 and check its status. For a description of the values returned, see
516 <a href="#performance_values">Performance Values</a> below.</p>
517 <p>The format in which the values are returned is similar to many other line based
518 protocols: Each value is printed on a separate line, each consisting of the
519 name of the value, a colon, one or more spaces and the actual value.</p>
520 <p>Example:</p>
521 <pre>
522  9 Statistics follow
523  QueueLength: 0
524  UpdatesReceived: 30
525  FlushesReceived: 2
526  UpdatesWritten: 13
527  DataSetsWritten: 390
528  TreeNodesNumber: 13
529  TreeDepth: 4
530  JournalBytes: 190
531  JournalRotate: 0</pre>
532 </dd>
533 <dt><strong><a name="update_filename_values_values" class="item"><strong>UPDATE</strong> <em>filename</em> <em>values</em> [<em>values</em> ...]</a></strong></dt>
535 <dd>
536 <p>Adds more data to a filename. This is <strong>the</strong> operation the daemon was designed
537 for, so describing the mechanism again is unnecessary. Read <a href="#how_it_works">HOW IT WORKS</a>
538 above for a detailed explanation.</p>
539 <p>Note that rrdcached only accepts absolute timestamps in the update values.
540 Updates strings like &quot;N:1:2:3&quot; are automatically converted to absolute
541 time by the RRD client library before sending to rrdcached.</p>
542 </dd>
543 <dt><strong><a name="wrote_filename" class="item"><strong>WROTE</strong> <em>filename</em></a></strong></dt>
545 <dd>
546 <p>This command is written to the journal after a file is successfully
547 written out to disk.  It is used during journal replay to determine which
548 updates have already been applied.  It is <em>only</em> valid in the journal; it
549 is not accepted from the other command channels.</p>
550 </dd>
551 <dt><strong><a name="batch" class="item"><strong>BATCH</strong></a></strong></dt>
553 <dd>
554 <p>This command initiates the bulk load of multiple commands.  This is
555 designed for installations with extremely high update rates, since it
556 permits more than one command to be issued per <code>read()</code> and <code>write()</code>.</p>
557 <p>All commands are executed just as they would be if given individually,
558 except for output to the user.  Messages indicating success are
559 suppressed, and error messages are delayed until the client is finished.</p>
560 <p>Command processing is finished when the client sends a dot (&quot;.&quot;) on its
561 own line.  After the client has finished, the server responds with an
562 error count and the list of error messages (if any).  Each error messages
563 indicates the number of the command to which it corresponds, and the error
564 message itself.  The first user command after <strong>BATCH</strong> is command number one.</p>
565 <pre>
566     client:  BATCH
567     server:  0 Go ahead.  End with dot '.' on its own line.
568     client:  UPDATE x.rrd 1223661439:1:2:3            &lt;--- command #1
569     client:  UPDATE y.rrd 1223661440:3:4:5            &lt;--- command #2
570     client:  and so on...
571     client:  .
572     server:  2 Errors
573     server:  1 message for command 1
574     server:  12 message for command 12</pre>
575 </dd>
576 <dt><strong><a name="quit" class="item"><strong>QUIT</strong></a></strong></dt>
578 <dd>
579 <p>Disconnect from rrdcached.</p>
580 </dd>
581 </dl>
582 <p>
583 </p>
584 <h2><a name="performance_values">Performance Values</a></h2>
585 <p>The following counters are returned by the <strong>STATS</strong> command:</p>
586 <dl>
587 <dt><strong><a name="queuelength" class="item"><strong>QueueLength</strong> <em>(unsigned 64bit integer)</em></a></strong></dt>
589 <dd>
590 <p>Number of nodes currently enqueued in the update queue.</p>
591 </dd>
592 <dt><strong><a name="updatesreceived" class="item"><strong>UpdatesReceived</strong> <em>(unsigned 64bit integer)</em></a></strong></dt>
594 <dd>
595 <p>Number of UPDATE commands received.</p>
596 </dd>
597 <dt><strong><a name="flushesreceived" class="item"><strong>FlushesReceived</strong> <em>(unsigned 64bit integer)</em></a></strong></dt>
599 <dd>
600 <p>Number of FLUSH commands received.</p>
601 </dd>
602 <dt><strong><a name="updateswritten" class="item"><strong>UpdatesWritten</strong> <em>(unsigned 64bit integer)</em></a></strong></dt>
604 <dd>
605 <p>Total number of updates, i.&nbsp;e. calls to <code>rrd_update_r</code>, since the
606 daemon was started.</p>
607 </dd>
608 <dt><strong><a name="datasetswritten" class="item"><strong>DataSetsWritten</strong> <em>(unsigned 64bit integer)</em></a></strong></dt>
610 <dd>
611 <p>Total number of &quot;data sets&quot; written to disk since the daemon was
612 started. A data set is one or more values passed to the <strong>UPDATE</strong>
613 command. For example: <code>1223661439:123:456</code> is one data set with two
614 values. The term &quot;data set&quot; is used to prevent confusion whether
615 individual values or groups of values are counted.</p>
616 </dd>
617 <dt><strong><a name="treenodesnumber" class="item"><strong>TreeNodesNumber</strong> <em>(unsigned 64bit integer)</em></a></strong></dt>
619 <dd>
620 <p>Number of nodes in the cache.</p>
621 </dd>
622 <dt><strong><a name="treedepth" class="item"><strong>TreeDepth</strong> <em>(unsigned 64bit integer)</em></a></strong></dt>
624 <dd>
625 <p>Depth of the tree used for fast key lookup.</p>
626 </dd>
627 <dt><strong><a name="journalbytes" class="item"><strong>JournalBytes</strong> <em>(unsigned 64bit integer)</em></a></strong></dt>
629 <dd>
630 <p>Total number of bytes written to the journal since startup.</p>
631 </dd>
632 <dt><strong><a name="journalrotate" class="item"><strong>JournalRotate</strong> <em>(unsigned 64bit integer)</em></a></strong></dt>
634 <dd>
635 <p>Number of times the journal has been rotated since startup.</p>
636 </dd>
637 </dl>
638 <p>
639 </p>
640 <hr />
641 <h1><a name="signals">SIGNALS</a></h1>
642 <dl>
643 <dt><strong><a name="sigint_and_sigterm" class="item">SIGINT and SIGTERM</a></strong></dt>
645 <dd>
646 <p>The daemon exits normally on receipt of either of these signals.  Pending
647 updates are handled in accordance with the <strong>-j</strong> and <strong>-F</strong> options.</p>
648 </dd>
649 <dt><strong><a name="sigusr1" class="item">SIGUSR1</a></strong></dt>
651 <dd>
652 <p>The daemon exits AFTER flushing all updates out to disk.  This may take a
653 while.</p>
654 </dd>
655 <dt><strong><a name="sigusr2" class="item">SIGUSR2</a></strong></dt>
657 <dd>
658 <p>The daemon exits immediately, without flushing updates out to disk.
659 Pending updates will be replayed from the journal when the daemon starts
660 up again.  <strong>WARNING: if journaling (-j) is NOT enabled, any pending
661 updates WILL BE LOST</strong>.</p>
662 </dd>
663 </dl>
664 <p>
665 </p>
666 <hr />
667 <h1><a name="bugs">BUGS</a></h1>
668 <p>No known bugs at the moment.</p>
669 <p>
670 </p>
671 <hr />
672 <h1><a name="see_also">SEE ALSO</a></h1>
673 <p><a href="././rrdtool.html">the rrdtool manpage</a>, <a href="././rrdgraph.html">the rrdgraph manpage</a></p>
674 <p>
675 </p>
676 <hr />
677 <h1><a name="author">AUTHOR</a></h1>
678 <p><strong>rrdcached</strong> and this manual page have been written by Florian Forster
679 &lt;octo&nbsp;at&nbsp;verplant.org&gt;.</p>
680 <p>
681 </p>
682 <hr />
683 <h1><a name="contributors">CONTRIBUTORS</a></h1>
684 <p>kevin brintnall &lt;<a href="mailto:kbrint@rufus.net">kbrint@rufus.net</a>&gt;</p>
686 </body>
688 </html>