1RRDCACHED(1) rrdtool RRDCACHED(1)
2
6 rrdcached - Data caching daemon for rrdtool
7
9 rrdcached [-P permissions] [-l address] [-s group] [-w timeout]
10 [-z delay] [-f timeout] [-p pid_file] [-t write_threads]
11 [-j journal_dir] [-F] [-g] [-b base_dir [-B]]
12
14 rrdcached is a daemon that receives updates to existing RRD files,
15 accumulates them and, if enough have been received or a defined time
16 has passed, writes the updates to the RRD file. A flush command may be
17 used to force writing of values to disk, so that graphing facilities
18 and similar can work with up-to-date data.
19 The daemon was written with big setups in mind. Those setups usually
21 run into IO related problems sooner or later for reasons that are
22 beyond the scope of this document. Check the wiki at the RRDtool
23 homepage for details. Also check "SECURITY CONSIDERATIONS" below before
24 using this daemon! A detailed description of how the daemon operates
25 can be found in the "HOW IT WORKS" section below.
26
28 -l address
29 Tells the daemon to bind to address and accept incoming connections
30 on that socket. If address begins with "unix:", everything
31 following that prefix is interpreted as the path to a UNIX domain
32 socket. Otherwise the address or node name are resolved using
33 "getaddrinfo()".
34 For network sockets, a port may be specified by using the form
36 "[address]:port". If the address is an IPv4 address or a fully
37 qualified domain name (i. e. the address contains at least one dot
38 (".")), the square brackets can be omitted, resulting in the
39 (simpler) "address:port" pattern. The default port is 42217/udp. If
40 you specify a network socket, it is mandatory to read the "SECURITY
41 CONSIDERATIONS" section.
42 The following formats are accepted. Please note that the address of
44 the UNIX domain socket must start with a slash in the second case!
45 unix:</path/to/unix.sock>
47 /<path/to/unix.sock>
48 <hostname-or-ip>
49 [<hostname-or-ip>]:<port>
50 <hostname-or-ipv4>:<port>
51 If the -l option is not specified the default address,
53 "unix:/tmp/rrdcached.sock", will be used.
54 -s group_name|gid
56 Set the group permissions of a UNIX domain socket. The option
57 accepts either a numeric group id or group name. That group will
58 then have both read and write permissions (the socket will have
59 file permissions 0750) for the socket and, therefore, is able to
60 send commands to the daemon. This may be useful in cases where you
61 cannot easily run all RRD processes with the same user privileges
62 (e.g. graph generating CGI scripts that typically run in the
63 permission context of the web server).
64 This option affects the following UNIX socket addresses (the
66 following -l options), i.e., you may specify different settings for
67 different sockets.
68 The default is not to change ownership or permissions of the socket
70 and, thus, use the system default.
71 -m mode
73 Set the file permissions of a UNIX domain socket. The option
74 accepts an octal number representing the bit pattern for the mode
75 (see chmod(1) for details).
76 Please note that not all systems honor this setting. On Linux,
78 read/write permissions are required to connect to a UNIX socket.
79 However, many BSD-derived systems ignore permissions for UNIX
80 sockets. See unix(7) for details.
81 This option affects the following UNIX socket addresses (the
83 following -l options), i.e., you may specify different settings for
84 different sockets.
85 The default is not to change ownership or permissions of the socket
87 and, thus, use the system default.
88 -P command[,command[,...]]
90 Specifies the commands accepted via a network socket. This allows
91 administrators of RRDCacheD to control the actions accepted from
92 various sources.
93 The arguments given to the -P option is a comma separated list of
95 commands. For example, to allow the "FLUSH" and "PENDING" commands
96 one could specify:
97 rrdcached -P FLUSH,PENDING $MORE_ARGUMENTS
99 The -P option affects the following socket addresses (the following
101 -l options). In the following example, only the IPv4 network socket
102 (address 10.0.0.1) will be restricted to the "FLUSH" and "PENDING"
103 commands:
104 rrdcached -l unix:/some/path -P FLUSH,PENDING -l 10.0.0.1
106 A complete list of available commands can be found in the section
108 "Valid Commands" below. There are two minor special exceptions:
109 · The "HELP" and "QUIT" commands are always allowed.
111 · If the "BATCH" command is accepted, the . command will
113 automatically be accepted, too.
114 Please also read "SECURITY CONSIDERATIONS" below.
116 -w timeout
118 Data is written to disk every timeout seconds. If this option is
119 not specified the default interval of 300 seconds will be used.
120 -z delay
122 If specified, rrdcached will delay writing of each RRD for a random
123 number of seconds in the range [0,delay). This will avoid too many
124 writes being queued simultaneously. This value should be no
125 greater than the value specified in -w. By default, there is no
126 delay.
127 -f timeout
129 Every timeout seconds the entire cache is searched for old values
130 which are written to disk. This only concerns files to which
131 updates have stopped, so setting this to a high value, such as
132 3600 seconds, is acceptable in most cases. This timeout defaults to
133 3600 seconds.
134 -p file
136 Sets the name and location of the PID-file. If not specified, the
137 default, "$localststedir/run/rrdcached.pid" will be used.
138 -t write_threads
140 Specifies the number of threads used for writing RRD files. The
141 default is 4. Increasing this number will allow rrdcached to have
142 more simultaneous I/O requests into the kernel. This may allow the
143 kernel to re-order disk writes, resulting in better disk
144 throughput.
145 -j dir
147 Write updates to a journal in dir. In the event of a program or
148 system crash, this will allow the daemon to write any updates that
149 were pending at the time of the crash.
150 On startup, the daemon will check for journal files in this
152 directory. If found, all updates therein will be read into memory
153 before the daemon starts accepting new connections.
154 The journal will be rotated with the same frequency as the flush
156 timer given by -f.
157 When journaling is enabled, the daemon will use a fast shutdown
159 procedure. Rather than flushing all files to disk, it will make
160 sure the journal is properly written and exit immediately.
161 Although the RRD data files are not fully up-to-date, no
162 information is lost; all pending updates will be replayed from the
163 journal next time the daemon starts up.
164 To disable fast shutdown, use the -F option.
166 -F ALWAYS flush all updates to the RRD data files when the daemon is
168 shut down, regardless of journal setting.
169 -g Run in the foreground. The daemon will not fork().
171 -b dir
173 The daemon will change into a specific directory at startup. All
174 files passed to the daemon, that are specified by a relative path,
175 will be interpreted to be relative to this directory. If not given
176 the default, "/tmp", will be used.
177 +------------------------+------------------------+
179 ! Command line ! File updated !
180 +------------------------+------------------------+
181 ! foo.rrd ! /tmp/foo.rrd !
182 ! foo/bar.rrd ! /tmp/foo/bar.rrd !
183 ! /var/lib/rrd/foo.rrd ! /var/lib/rrd/foo.rrd !
184 +------------------------+------------------------+
185 Paths given on the command line and paths actually
186 updated by the daemon, assuming the base directory
187 "/tmp".
188 WARNING: The paths up to and including the base directory MUST NOT
190 BE symbolic links. In other words, if the base directory is
191 specified as:
192 -b /base/dir/somewhere
194 ... then NONE of the following should be symbolic links:
196 /base
198 /base/dir
199 /base/dir/somewhere
200 -B Only permit writes into the base directory specified in -b (and any
202 sub-directories). This does NOT detect symbolic links. Paths
203 containing "../" will also be blocked.
204
206 The following commands may be made aware of the rrdcached using the
207 command line argument --daemon or the environment variable
208 RRDCACHED_ADDRESS:
209 · dump
211 · fetch
213 · flush
215 · graph
217 · graphv
219 · info
221 · last
223 · lastupdate
225 · update
227 · xport
229 The update command can send values to the daemon instead of writing
231 them to the disk itself. All other commands can send a FLUSH command
232 (see below) to the daemon before accessing the files, so they work with
233 up-to-date data even if the cache timeout is large.
234
236 The daemon reports errors in one of two ways: During startup, error
237 messages are printed to "STDERR". One of the steps when starting up is
238 to fork to the background and closing "STDERR" - after this writing
239 directly to the user is no longer possible. Once this has happened, the
240 daemon will send log messages to the system logging daemon using
241 syslog(3). The facility used is "LOG_DAEMON".
242
244 When receiving an update, rrdcached does not write to disk but looks
245 for an entry for that file in its internal tree. If not found, an entry
246 is created including the current time (called "First" in the diagram
247 below). This time is not the time specified on the command line but the
248 time the operating system considers to be "now". The value and time of
249 the value (called "Time" in the diagram below) are appended to the tree
250 node.
251 When appending a value to a tree node, it is checked whether it's time
253 to write the values to disk. Values are written to disk if
254 "now() - First >= timeout", where "timeout" is the timeout specified
255 using the -w option, see "OPTIONS". If the values are "old enough" they
256 will be enqueued in the "update queue", i. e. they will be appended to
257 the linked list shown below. Because the tree nodes and the elements
258 of the linked list are the same data structures in memory, any update
259 to a file that has already been enqueued will be written with the next
260 write to the RRD file, too.
261 A separate "update thread" constantly dequeues the first element in the
263 update queue and writes all its values to the appropriate file. So as
264 long as the update queue is not empty files are written at the highest
265 possible rate.
266 Since the timeout of files is checked only when new values are added to
268 the file, "dead" files, i. e. files that are not updated anymore, would
269 never be written to disk. Therefore, every now and then, controlled by
270 the -f option, the entire tree is walked and all "old" values are
271 enqueued. Since this only affects "dead" files and walking the tree is
272 relatively expensive, you should set the "flush interval" to a
273 reasonably high value. The default is 3600 seconds (one hour).
274 The downside of caching values is that they won't show up in graphs
276 generated from the RRD files. To get around this, the daemon provides
277 the "flush command" to flush specific files. This means that the file
278 is inserted at the head of the update queue or moved there if it is
279 already enqueued. The flush command will return only after the file's
280 pending updates have been written to disk.
281 +------+ +------+ +------+
283 ! head ! ! root ! ! tail !
284 +---+--+ +---+--+ +---+--+
285 ! /\ !
286 ! / \ !
287 ! /\ /\ !
288 ! /\/\ \ `----------------- ... --------, !
289 V / `-------, ! V
290 +---+----+---+ +------+-----+ +---+----+---+
291 ! File: foo ! ! File: bar ! ! File: qux !
292 ! First: 101 ! ! First: 119 ! ! First: 180 !
293 ! Next:&bar -+--->! Next:&... -+---> ... --->! Next:NULL !
294 | Prev:NULL !<---+-Prev:&foo !<--- ... ----+-Prev: &... !
295 +============+ +============+ +============+
296 ! Time: 100 ! ! Time: 120 ! ! Time: 180 !
297 ! Value: 10 ! ! Value: 0.1 ! ! Value: 2,2 !
298 +------------+ +------------+ +------------+
299 ! Time: 110 ! ! Time: 130 ! ! Time: 190 !
300 ! Value: 26 ! ! Value: 0.1 ! ! Value: 7,3 !
301 +------------+ +------------+ +------------+
302 : : : : : :
303 +------------+ +------------+ +------------+
304 ! Time: 230 ! ! Time: 250 ! ! Time: 310 !
305 ! Value: 42 ! ! Value: 0.2 ! ! Value: 1,2 !
306 +------------+ +------------+ +------------+
307 The above diagram demonstrates:
309 · Files/values are stored in a (balanced) tree.
311 · Tree nodes and entries in the update queue are the same data
313 structure.
314 · The local time ("First") and the time specified in updates ("Time")
316 may differ.
317 · Timed out values are inserted at the "tail".
319 · Explicitly flushed values are inserted at the "head".
321 · ASCII art rocks.
323
325 Authentication
326 There is no authentication.
327 The client/server protocol does not yet have any authentication
329 mechanism. It is likely that authentication and encryption will be
330 added in a future version, but for the time being it is the
331 administrator's responsibility to secure the traffic from/to the
332 daemon!
333 It is highly recommended to install a packet filter or similar
335 mechanism to prevent unauthorized connections. Unless you have a
336 dedicated VLAN or VPN for this, using network sockets is probably a bad
337 idea!
338 Authorization
340 There is minimal per-socket authorization.
341 Authorization is currently done on a per-socket basis. That means each
343 socket has a list of commands it will accept and it will accept. It
344 will accept only those commands explicitly listed but it will
345 (currently) accept these commands from anyone reaching the socket.
346 If the networking sockets are to be used, it is necessary to restrict
348 the accepted commands to those needed by external clients. If, for
349 example, external clients want to draw graphs of the cached data, they
350 should only be allowed to use the "FLUSH" command.
351 Encryption
353 There is no encryption.
354 Again, this may be added in the future, but for the time being it is
356 your job to keep your private data private. Install a VPN or an
357 encrypted tunnel if you statistics are confidential!
358 Sanity checking
360 There is no sanity checking.
361 The daemon will blindly write to any file it gets told, so you really
363 should create a separate user just for this daemon. Also it does not do
364 any sanity checks, so if it gets told to write values for a time far in
365 the future, your files will be messed up good!
366 Conclusion
368 · Security is the job of the administrator.
369 · We recommend to allow write access via UNIX domain sockets only.
371 · You have been warned.
373
375 The daemon communicates with clients using a line based ASCII protocol
376 which is easy to read and easy to type. This makes it easy for scripts
377 to implement the protocol and possible for users to use telnet to
378 connect to the daemon and test stuff "by hand".
379 The protocol is line based, this means that each record consists of one
381 or more lines. A line is terminated by the line feed character 0x0A,
382 commonly written as "\n". In the examples below, this character will be
383 written as "<LF>" ("line feed").
384 After the connection has been established, the client is expected to
386 send a "command". A command consists of the command keyword, possibly
387 some arguments, and a terminating newline character. For a list of
388 commands, see "Valid Commands" below.
389 Example:
391 FLUSH /tmp/foo.rrd<LF>
393 The daemon answers with a line consisting of a status code and a short
395 status message, separated by one or more space characters. A negative
396 status code signals an error, a positive status code or zero signal
397 success. If the status code is greater than zero, it indicates the
398 number of lines that follow the status line.
399 Examples:
401 0 Success<LF>
403 2 Two lines follow<LF>
405 This is the first line<LF>
406 And this is the second line<LF>
407 Valid Commands
409 The following commands are understood by the daemon:
410 FLUSH filename
412 Causes the daemon to put filename to the head of the update queue
413 (possibly moving it there if the node is already enqueued). The
414 answer will be sent after the node has been dequeued.
415 FLUSHALL
417 Causes the daemon to start flushing ALL pending values to disk.
418 This returns immediately, even though the writes may take a long
419 time.
420 PENDING filename
422 Shows any "pending" updates for a file, in order. The updates
423 shown have not yet been written to the underlying RRD file.
424 FORGET filename
426 Removes filename from the cache. Any pending updates WILL BE LOST.
427 QUEUE
429 Shows the files that are on the output queue. Returns zero or more
430 lines in the following format, where <num_vals> is the number of
431 values to be written for the <file>:
432 <num_vals> <file>
434 HELP [command]
436 Returns a short usage message. If no command is given, or command
437 is HELP, a list of commands supported by the daemon is returned.
438 Otherwise a short description, possibly containing a pointer to a
439 manual page, is returned. Obviously, this is meant for interactive
440 usage and the format in which the commands and usage summaries are
441 returned is not well defined.
442 STATS
444 Returns a list of metrics which can be used to measure the daemons
445 performance and check its status. For a description of the values
446 returned, see "Performance Values" below.
447 The format in which the values are returned is similar to many
449 other line based protocols: Each value is printed on a separate
450 line, each consisting of the name of the value, a colon, one or
451 more spaces and the actual value.
452 Example:
454 9 Statistics follow
456 QueueLength: 0
457 UpdatesReceived: 30
458 FlushesReceived: 2
459 UpdatesWritten: 13
460 DataSetsWritten: 390
461 TreeNodesNumber: 13
462 TreeDepth: 4
463 JournalBytes: 190
464 JournalRotate: 0
465 UPDATE filename values [values ...]
467 Adds more data to a filename. This is the operation the daemon was
468 designed for, so describing the mechanism again is unnecessary.
469 Read "HOW IT WORKS" above for a detailed explanation.
470 Note that rrdcached only accepts absolute timestamps in the update
472 values. Updates strings like "N:1:2:3" are automatically converted
473 to absolute time by the RRD client library before sending to
474 rrdcached.
475 WROTE filename
477 This command is written to the journal after a file is successfully
478 written out to disk. It is used during journal replay to determine
479 which updates have already been applied. It is only valid in the
480 journal; it is not accepted from the other command channels.
481 BATCH
483 This command initiates the bulk load of multiple commands. This is
484 designed for installations with extremely high update rates, since
485 it permits more than one command to be issued per read() and
486 write().
487 All commands are executed just as they would be if given
489 individually, except for output to the user. Messages indicating
490 success are suppressed, and error messages are delayed until the
491 client is finished.
492 Command processing is finished when the client sends a dot (".") on
494 its own line. After the client has finished, the server responds
495 with an error count and the list of error messages (if any). Each
496 error messages indicates the number of the command to which it
497 corresponds, and the error message itself. The first user command
498 after BATCH is command number one.
499 client: BATCH
501 server: 0 Go ahead. End with dot '.' on its own line.
502 client: UPDATE x.rrd 1223661439:1:2:3 <--- command #1
503 client: UPDATE y.rrd 1223661440:3:4:5 <--- command #2
504 client: and so on...
505 client: .
506 server: 2 Errors
507 server: 1 message for command 1
508 server: 12 message for command 12
509 QUIT
511 Disconnect from rrdcached.
512 Performance Values
514 The following counters are returned by the STATS command:
515 QueueLength (unsigned 64bit integer)
517 Number of nodes currently enqueued in the update queue.
518 UpdatesReceived (unsigned 64bit integer)
520 Number of UPDATE commands received.
521 FlushesReceived (unsigned 64bit integer)
523 Number of FLUSH commands received.
524 UpdatesWritten (unsigned 64bit integer)
526 Total number of updates, i. e. calls to "rrd_update_r", since the
527 daemon was started.
528 DataSetsWritten (unsigned 64bit integer)
530 Total number of "data sets" written to disk since the daemon was
531 started. A data set is one or more values passed to the UPDATE
532 command. For example: "1223661439:123:456" is one data set with two
533 values. The term "data set" is used to prevent confusion whether
534 individual values or groups of values are counted.
535 TreeNodesNumber (unsigned 64bit integer)
537 Number of nodes in the cache.
538 TreeDepth (unsigned 64bit integer)
540 Depth of the tree used for fast key lookup.
541 JournalBytes (unsigned 64bit integer)
543 Total number of bytes written to the journal since startup.
544 JournalRotate (unsigned 64bit integer)
546 Number of times the journal has been rotated since startup.
547
549 SIGINT and SIGTERM
550 The daemon exits normally on receipt of either of these signals.
551 Pending updates are handled in accordance with the -j and -F
552 options.
553 SIGUSR1
555 The daemon exits AFTER flushing all updates out to disk. This may
556 take a while.
557 SIGUSR2
559 The daemon exits immediately, without flushing updates out to disk.
560 Pending updates will be replayed from the journal when the daemon
561 starts up again. WARNING: if journaling (-j) is NOT enabled, any
562 pending updates WILL BE LOST.
563
565 No known bugs at the moment.
566
568 rrdtool, rrdgraph
569
571 Florian Forster <octo at verplant.org>
572 Both rrdcached and this manual page have been written by Florian.
574
576 kevin brintnall <kbrint@rufus.net>
5771.4.4 2010-03-22 RRDCACHED(1)