1RRDCACHED(1) rrdtool RRDCACHED(1)
2
6 rrdcached - Data caching daemon for rrdtool
7
9 rrdcached [-a alloc_size] [-b base_dir [-B]] [-F] [-f timeout]
10 [-G group]] [-g] [-j journal_dir] [-L] [-l address] [-m mode] [-O]
11 [-P permissions] [-p pid_file] [-R] [-s group] [-t write_threads]
12 [-U user]] [-V log_level] [-w timeout] [-z delay]
13
15 rrdcached is a daemon that receives updates to existing RRD files,
16 accumulates them and, if enough have been received or a defined time
17 has passed, writes the updates to the RRD file. A flush command may be
18 used to force writing of values to disk, so that graphing facilities
19 and similar can work with up-to-date data.
20 The daemon was written with big setups in mind. Those setups usually
22 run into IO related problems sooner or later for reasons that are
23 beyond the scope of this document. Check the wiki at the RRDtool
24 homepage for details. Also check "SECURITY CONSIDERATIONS" below before
25 using this daemon! A detailed description of how the daemon operates
26 can be found in the "HOW IT WORKS" section below.
27
29 -l address
30 Tells the daemon to bind to address and accept incoming TCP
31 connections on that socket. If address begins with "unix:",
32 everything following that prefix is interpreted as the path to a
33 UNIX domain socket. Otherwise the address or node name are resolved
34 using "getaddrinfo()".
35 For network sockets, a port may be specified by using the form
37 "[address]:port". If the address is an IPv4 address or a fully
38 qualified domain name (i. e. the address contains at least one dot
39 (".")), the square brackets can be omitted, resulting in the
40 (simpler) "address:port" pattern. The default port is 42217. If you
41 specify a network socket, it is mandatory to read the "SECURITY
42 CONSIDERATIONS" section.
43 The following formats are accepted. Please note that the address of
45 the UNIX domain socket must start with a slash in the second case!
46 unix:</path/to/unix.sock>
48 /<path/to/unix.sock>
49 <hostname-or-ip>
50 [<hostname-or-ip>]:<port>
51 <hostname-or-ipv4>:<port>
52 Given a port without a host (e.g. "-l :42217") the daemon will
54 listen on that port on all network interfaces. Use "-L" to avoid
55 the need to explicitly provide the port if the default port is
56 desired.
57 If no -l option is not specified the default address,
59 "unix:/tmp/rrdcached.sock", will be used. Multiple -l options may
60 be provided.
61 -L Tells the daemon to bind to the default TCP port on all available
63 interfaces. It is equivalent to "-l ''" without the confusion of
64 the empty string parameter.
65 -s group_name|gid
67 Set the group permissions of a UNIX domain socket. The option
68 accepts either a numeric group id or group name. That group will
69 then have both read and write permissions (the socket will have
70 file permissions 0760) for the socket and, therefore, is able to
71 send commands to the daemon. This may be useful in cases where you
72 cannot easily run all RRD processes with the same user privileges
73 (e.g. graph generating CGI scripts that typically run in the
74 permission context of the web server).
75 This option affects the following UNIX socket addresses (the
77 following -l options) or the default socket (if no -l options have
78 been specified), i.e., you may specify different settings for
79 different sockets.
80 The default is not to change ownership or permissions of the socket
82 and, thus, use the system default.
83 -m mode
85 Set the file permissions of a UNIX domain socket. The option
86 accepts an octal number representing the bit pattern for the mode
87 (see chmod(1) for details).
88 Please note that not all systems honor this setting. On Linux,
90 read/write permissions are required to connect to a UNIX socket.
91 However, many BSD-derived systems ignore permissions for UNIX
92 sockets. See unix(7) for details.
93 This option affects the following UNIX socket addresses (the
95 following -l options) or the default socket (if no -l options have
96 been specified), i.e., you may specify different settings for
97 different sockets.
98 The default is not to change ownership or permissions of the socket
100 and, thus, use the system default.
101 -P command[,command[,...]]
103 Specifies the commands accepted via both a network and a UNIX
104 socket. This allows administrators of RRDCacheD to control the
105 actions accepted from various sources.
106 The arguments given to the -P option is a comma separated list of
108 commands. For example, to allow the "FLUSH" and "PENDING" commands
109 one could specify:
110 rrdcached -P FLUSH,PENDING $MORE_ARGUMENTS
112 The -P option affects the following socket addresses (the following
114 -l options) or the default socket (if no -l options have been
115 specified). In the following example, only the IPv4 network socket
116 (address 10.0.0.1) will be restricted to the "FLUSH" and "PENDING"
117 commands:
118 rrdcached -l unix:/some/path -P FLUSH,PENDING -l 10.0.0.1
120 A complete list of available commands can be found in the section
122 "Valid Commands" below. There are two minor special exceptions:
123 · The "HELP" and "QUIT" commands are always allowed.
125 · If the "BATCH" command is accepted, the . command will
127 automatically be accepted, too.
128 Please also read "SECURITY CONSIDERATIONS" below.
130 -V log_level
132 rrdcached under load can severely flood the logs. This command line
133 option specifies the maximum log_level to be used, meaning that a
134 message with verbosity higher than log_level is muted (LOG_EMERG
135 being the lowest and LOG_DEBUG highest).
136 Accepted values for "log_level" (lowest to highest verbosity):
138 LOG_EMERG, LOG_ALERT, LOG_CRIT, LOG_ERR, LOG_WARNING, LOG_NOTICE,
139 LOG_INFO, LOG_DEBUG
140 Default log level when this flag is NOT present: LOG_ERR
142 See also: syslog.h
144 -w timeout
146 Data is written to disk every timeout seconds. An optional suffix
147 may be used (e.g. "5m" instead of 300 seconds). If this option is
148 not specified the default interval of 300 seconds will be used.
149 -z delay
151 If specified, rrdcached will delay writing of each RRD for a random
152 number of seconds in the range [0,delay). This will avoid too many
153 writes being queued simultaneously. This value should be no
154 greater than the value specified in -w. An optional suffix may be
155 used (e.g. "3m" instead of 180 seconds). By default, there is no
156 delay.
157 -f timeout
159 Every timeout seconds the entire cache is searched for old values
160 which are written to disk. This only concerns files to which
161 updates have stopped, so setting this to a high value, such as
162 3600 seconds, is acceptable in most cases. An optional suffix may
163 be used (e.g. "1h" instead of 3600 seconds). This timeout defaults
164 to 3600 seconds.
165 -p file
167 Sets the name and location of the PID-file. If not specified, the
168 default, "$localstatedir/run/rrdcached.pid" will be used.
169 -t write_threads
171 Specifies the number of threads used for writing RRD files. The
172 default is 4. Increasing this number will allow rrdcached to have
173 more simultaneous I/O requests into the kernel. This may allow the
174 kernel to re-order disk writes, resulting in better disk
175 throughput.
176 -j dir
178 Write updates to a journal in dir. In the event of a program or
179 system crash, this will allow the daemon to write any updates that
180 were pending at the time of the crash.
181 On startup, the daemon will check for journal files in this
183 directory. If found, all updates therein will be read into memory
184 before the daemon starts accepting new connections.
185 The journal will be rotated with the same frequency as the flush
187 timer given by -f.
188 When journaling is enabled, the daemon will use a fast shutdown
190 procedure. Rather than flushing all files to disk, it will make
191 sure the journal is properly written and exit immediately.
192 Although the RRD data files are not fully up-to-date, no
193 information is lost; all pending updates will be replayed from the
194 journal next time the daemon starts up.
195 To disable fast shutdown, use the -F option.
197 -F ALWAYS flush all updates to the RRD data files when the daemon is
199 shut down, regardless of journal setting.
200 -g Run in the foreground. The daemon will not fork().
202 -b dir
204 The daemon will change into a specific directory at startup. All
205 files passed to the daemon, that are specified by a relative path,
206 will be interpreted to be relative to this directory. If not given
207 the default, "/tmp", will be used.
208 +------------------------+------------------------+
210 ! Command line ! File updated !
211 +------------------------+------------------------+
212 ! foo.rrd ! /tmp/foo.rrd !
213 ! foo/bar.rrd ! /tmp/foo/bar.rrd !
214 ! /var/lib/rrd/foo.rrd ! /var/lib/rrd/foo.rrd !
215 +------------------------+------------------------+
216 Paths given on the command line and paths actually
217 updated by the daemon, assuming the base directory
218 "/tmp".
219 WARNING: The paths up to and including the base directory MUST NOT
221 BE symbolic links. In other words, if the base directory is
222 specified as:
223 -b /base/dir/somewhere
225 ... then NONE of the following should be symbolic links:
227 /base
229 /base/dir
230 /base/dir/somewhere
231 -B Only permit writes into the base directory specified in -b (and any
233 sub-directories). This does NOT detect symbolic links. Paths
234 containing "../" will also be blocked.
235 -R Permit recursive subdirectory creation in the base directory
237 specified in -b (and any sub-directories). Can only be used when -B
238 is also set.
239 -a alloc_size
241 Allocate value pointers in chunks of alloc_size. This may improve
242 CPU utilization on machines with slow "realloc()" implementations,
243 in exchange for slightly higher memory utilization. The default
244 is 1. Do not set this more than the -w value divided by your
245 average RRD step size.
246 -O Prevent the CREATE command from overwriting existing files, even
248 when it is instructed to do so. This is for added security.
249 -G -group
251 When running as daemon and invoked from a privileged account, reset
252 group privileges to those of group. The group may be specified as
253 a name or as a group ID. The daemon will exit with a diagnostic if
254 it cannot successfully transition to the specified group.
255 -U -user
257 When running as daemon and invoked from a privileged account, reset
258 user privileges to those of user. The user may be specified as a
259 name or as a user ID. The daemon will exit with a diagnostic if it
260 cannot successfully transition to the specified user.
261
263 The following commands may be made aware of the rrdcached using the
264 command line argument --daemon or the environment variable
265 RRDCACHED_ADDRESS:
266 · dump
268 · fetch
270 · flush
272 · graph
274 · graphv
276 · info
278 · first
280 · last
282 · lastupdate
284 · update
286 · xport
288 · create
290 · list
292 The update command can send values to the daemon instead of writing
294 them to the disk itself. All other commands can send a FLUSH command
295 (see below) to the daemon before accessing the files, so they work with
296 up-to-date data even if the cache timeout is large.
297
299 The daemon reports errors in one of two ways: During startup, error
300 messages are printed to "STDERR". One of the steps when starting up is
301 to fork to the background and closing "STDERR" - after this writing
302 directly to the user is no longer possible. Once this has happened, the
303 daemon will send log messages to the system logging daemon using
304 syslog(3). The facility used is "LOG_DAEMON".
305
307 When receiving an update, rrdcached does not write to disk but looks
308 for an entry for that file in its internal tree. If not found, an entry
309 is created including the current time (called "First" in the diagram
310 below). This time is not the time specified on the command line but the
311 time the operating system considers to be "now". The value and time of
312 the value (called "Time" in the diagram below) are appended to the tree
313 node.
314 When appending a value to a tree node, it is checked whether it's time
316 to write the values to disk. Values are written to disk if
317 "now() - First >= timeout", where "timeout" is the timeout specified
318 using the -w option, see "OPTIONS". If the values are "old enough" they
319 will be enqueued in the "update queue", i. e. they will be appended to
320 the linked list shown below. Because the tree nodes and the elements
321 of the linked list are the same data structures in memory, any update
322 to a file that has already been enqueued will be written with the next
323 write to the RRD file, too.
324 A separate "update thread" constantly dequeues the first element in the
326 update queue and writes all its values to the appropriate file. So as
327 long as the update queue is not empty files are written at the highest
328 possible rate.
329 Since the timeout of files is checked only when new values are added to
331 the file, "dead" files, i. e. files that are not updated anymore, would
332 never be written to disk. Therefore, every now and then, controlled by
333 the -f option, the entire tree is walked and all "old" values are
334 enqueued. Since this only affects "dead" files and walking the tree is
335 relatively expensive, you should set the "flush interval" to a
336 reasonably high value. The default is 3600 seconds (one hour).
337 The downside of caching values is that they won't show up in graphs
339 generated from the RRD files. To get around this, the daemon provides
340 the "flush command" to flush specific files. This means that the file
341 is inserted at the head of the update queue or moved there if it is
342 already enqueued. The flush command will return only after the file's
343 pending updates have been written to disk.
344 +------+ +------+ +------+
346 ! head ! ! root ! ! tail !
347 +---+--+ +---+--+ +---+--+
348 ! /\ !
349 ! / \ !
350 ! /\ /\ !
351 ! /\/\ \ `----------------- ... --------, !
352 V / `-------, ! V
353 +---+----+---+ +------+-----+ +---+----+---+
354 ! File: foo ! ! File: bar ! ! File: qux !
355 ! First: 101 ! ! First: 119 ! ! First: 180 !
356 ! Next:&bar -+--->! Next:&... -+---> ... --->! Next:NULL !
357 | Prev:NULL !<---+-Prev:&foo !<--- ... ----+-Prev: &... !
358 +============+ +============+ +============+
359 ! Time: 100 ! ! Time: 120 ! ! Time: 180 !
360 ! Value: 10 ! ! Value: 0.1 ! ! Value: 2,2 !
361 +------------+ +------------+ +------------+
362 ! Time: 110 ! ! Time: 130 ! ! Time: 190 !
363 ! Value: 26 ! ! Value: 0.1 ! ! Value: 7,3 !
364 +------------+ +------------+ +------------+
365 : : : : : :
366 +------------+ +------------+ +------------+
367 ! Time: 230 ! ! Time: 250 ! ! Time: 310 !
368 ! Value: 42 ! ! Value: 0.2 ! ! Value: 1,2 !
369 +------------+ +------------+ +------------+
370 The above diagram demonstrates:
372 · Files/values are stored in a (balanced) tree.
374 · Tree nodes and entries in the update queue are the same data
376 structure.
377 · The local time ("First") and the time specified in updates ("Time")
379 may differ.
380 · Timed out values are inserted at the "tail".
382 · Explicitly flushed values are inserted at the "head".
384 · ASCII art rocks.
386
388 Authentication
389 If your rrdtool installation was built without libwrap there is no form
390 of authentication for clients connecting to the rrdcache daemon!
391 If your rrdtool installation was built with libwrap then you can use
393 hosts_access to restrict client access to the rrdcache daemon
394 (rrdcached). For more information on how to use hosts_access to
395 restrict access to the rrdcache daemon you should read the
396 hosts_access(5) man pages.
397 It is still highly recommended to install a packet filter or similar
399 mechanism to prevent unauthorized connections. Unless you have a
400 dedicated VLAN or VPN for this, using network sockets is probably a bad
401 idea!
402 Authorization
404 There is minimal per-socket authorization.
405 Authorization is currently done on a per-socket basis. That means each
407 socket has a list of commands it will accept and it will accept. It
408 will accept only those commands explicitly listed but it will
409 (currently) accept these commands from anyone reaching the socket.
410 If the networking sockets are to be used, it is necessary to restrict
412 the accepted commands to those needed by external clients. If, for
413 example, external clients want to draw graphs of the cached data, they
414 should only be allowed to use the "FLUSH" command.
415 Authorization does not work when rrcached is socket-activated by
417 systemd.
418 Encryption
420 There is no encryption.
421 Again, this may be added in the future, but for the time being it is
423 your job to keep your private data private. Install a VPN or an
424 encrypted tunnel if you statistics are confidential!
425 Sanity checking
427 There is no sanity checking.
428 The daemon will blindly write to any file it gets told, so you really
430 should create a separate user just for this daemon. Also it does not do
431 any sanity checks, so if it gets told to write values for a time far in
432 the future, your files will be messed up good!
433 Conclusion
435 · Security is the job of the administrator.
436 · We recommend to allow write access via UNIX domain sockets only.
438 · You have been warned.
440
442 The daemon communicates with clients using a line based ASCII protocol
443 which is easy to read and easy to type. This makes it easy for scripts
444 to implement the protocol and possible for users to use telnet to
445 connect to the daemon and test stuff "by hand".
446 The protocol is line based, this means that each record consists of one
448 or more lines. A line is terminated by the line feed character 0x0A,
449 commonly written as "\n". In the examples below, this character will be
450 written as "<LF>" ("line feed").
451 After the connection has been established, the client is expected to
453 send a "command". A command consists of the command keyword, possibly
454 some arguments, and a terminating newline character. For a list of
455 commands, see "Valid Commands" below.
456 Example:
458 FLUSH /tmp/foo.rrd<LF>
460 The daemon answers with a line consisting of a status code and a short
462 status message, separated by one or more space characters. A negative
463 status code signals an error, a positive status code or zero signal
464 success. If the status code is greater than zero, it indicates the
465 number of lines that follow the status line.
466 Examples:
468 0 Success<LF>
470 2 Two lines follow<LF>
472 This is the first line<LF>
473 And this is the second line<LF>
474 Valid Commands
476 The following commands are understood by the daemon:
477 FLUSH filename
479 Causes the daemon to put filename to the head of the update queue
480 (possibly moving it there if the node is already enqueued). The
481 answer will be sent after the node has been dequeued.
482 FLUSHALL
484 Causes the daemon to start flushing ALL pending values to disk.
485 This returns immediately, even though the writes may take a long
486 time.
487 PENDING filename
489 Shows any "pending" updates for a file, in order. The updates
490 shown have not yet been written to the underlying RRD file.
491 FETCH filename CF [start [end] [ds ...]]
493 Calls "rrd_fetch" with the specified arguments and returns the
494 result in text form. If necessary, the file is flushed to disk
495 first. The client side function "rrdc_fetch" (declared in
496 "rrd_client.h") parses the output and behaves just like
497 "rrd_fetch_r" for easy integration of remote queries. ds defines
498 the columns to dump - if none are given then all are returned
499 FETCHBIN filename CF [start [end] [ds ...]]
501 Calls "rrd_fetch" with the specified arguments and returns the
502 result in text/binary form to avoid unnecessary un/marshalling
503 overhead. If necessary, the file is flushed to disk first. The
504 client side function "rrdc_fetch" (declared in "rrd_client.h")
505 parses the output and behaves just like "rrd_fetch_r" for easy
506 integration of remote queries. ds defines the columns to dump - if
507 none are given then all are returned
508 FORGET filename
510 Removes filename from the cache. Any pending updates WILL BE LOST.
511 QUEUE
513 Shows the files that are on the output queue. Returns zero or more
514 lines in the following format, where <num_vals> is the number of
515 values to be written for the <file>:
516 <num_vals> <file>
518 HELP [command]
520 Returns a short usage message. If no command is given, or command
521 is HELP, a list of commands supported by the daemon is returned.
522 Otherwise a short description, possibly containing a pointer to a
523 manual page, is returned. Obviously, this is meant for interactive
524 usage and the format in which the commands and usage summaries are
525 returned is not well defined.
526 STATS
528 Returns a list of metrics which can be used to measure the daemons
529 performance and check its status. For a description of the values
530 returned, see "Performance Values" below.
531 The format in which the values are returned is similar to many
533 other line based protocols: Each value is printed on a separate
534 line, each consisting of the name of the value, a colon, one or
535 more spaces and the actual value.
536 Example:
538 9 Statistics follow
540 QueueLength: 0
541 UpdatesReceived: 30
542 FlushesReceived: 2
543 UpdatesWritten: 13
544 DataSetsWritten: 390
545 TreeNodesNumber: 13
546 TreeDepth: 4
547 JournalBytes: 190
548 JournalRotate: 0
549 PING
551 PING-PONG, this is very useful when using connection pool between
552 user client and RRDCACHED.
553 Example:
555 0 PONG
557 UPDATE filename values [values ...]
559 Adds more data to a filename. This is the operation the daemon was
560 designed for, so describing the mechanism again is unnecessary.
561 Read "HOW IT WORKS" above for a detailed explanation.
562 Note that rrdcached only accepts absolute timestamps in the update
564 values. Updates strings like "N:1:2:3" are automatically converted
565 to absolute time by the RRD client library before sending to
566 rrdcached.
567 WROTE filename
569 This command is written to the journal after a file is successfully
570 written out to disk. It is used during journal replay to determine
571 which updates have already been applied. It is only valid in the
572 journal; it is not accepted from the other command channels.
573 FIRST filename [rranum]
575 Return the timestamp for the first CDP in the specified RRA.
576 Default is to use RRA zero if none is specified.
577 LAST filename
579 Return the timestamp for the last update to the specified RRD. Note
580 that the cache is not flushed before checking, as the client is
581 expected to request this separately if it is required.
582 INFO filename
584 Return the configuration information for the specified RRD. Note
585 that the cache is not flushed before checking, as the client is
586 expected to request this separately if it is required.
587 The information is returned, one item per line, with the format:
589 I<keyname> I<type> I<value>
591 CREATE filename [-s stepsize] [-b begintime] [-O] DSdefinitions ...
593 RRAdefinitions ...
594 This will create the RRD file according to the supplied parameters,
595 provided the parameters are valid, and (if the -O option is given
596 or if the rrdcached was started with the -O flag) the specified
597 filename does not already exist.
598 BATCH
600 This command initiates the bulk load of multiple commands. This is
601 designed for installations with extremely high update rates, since
602 it permits more than one command to be issued per read() and
603 write().
604 All commands are executed just as they would be if given
606 individually, except for output to the user. Messages indicating
607 success are suppressed, and error messages are delayed until the
608 client is finished.
609 Command processing is finished when the client sends a dot (".") on
611 its own line. After the client has finished, the server responds
612 with an error count and the list of error messages (if any). Each
613 error messages indicates the number of the command to which it
614 corresponds, and the error message itself. The first user command
615 after BATCH is command number one.
616 client: BATCH
618 server: 0 Go ahead. End with dot '.' on its own line.
619 client: UPDATE x.rrd 1223661439:1:2:3 <--- command #1
620 client: UPDATE y.rrd 1223661440:3:4:5 <--- command #2
621 client: and so on...
622 client: .
623 server: 2 Errors
624 server: 1 message for command 1
625 server: 12 message for command 12
626 LIST [RECURSIVE] I/<path>
628 This command allows to list directories and rrd databases as seen
629 by the daemon. The root "directory" is the base_dir (see '-b
630 dir'). When invoked with 'LIST RECURSIVE /<path>' it will behave
631 similarly to 'ls -R' but limited to rrd files (listing all the rrd
632 bases in the subtree of <path>, skipping empty directories).
633 QUIT
635 Disconnect from rrdcached.
636 Performance Values
638 The following counters are returned by the STATS command:
639 QueueLength (unsigned 64bit integer)
641 Number of nodes currently enqueued in the update queue.
642 UpdatesReceived (unsigned 64bit integer)
644 Number of UPDATE commands received.
645 FlushesReceived (unsigned 64bit integer)
647 Number of FLUSH commands received.
648 UpdatesWritten (unsigned 64bit integer)
650 Total number of updates, i. e. calls to "rrd_update_r", since the
651 daemon was started.
652 DataSetsWritten (unsigned 64bit integer)
654 Total number of "data sets" written to disk since the daemon was
655 started. A data set is one or more values passed to the UPDATE
656 command. For example: "1223661439:123:456" is one data set with two
657 values. The term "data set" is used to prevent confusion whether
658 individual values or groups of values are counted.
659 TreeNodesNumber (unsigned 64bit integer)
661 Number of nodes in the cache.
662 TreeDepth (unsigned 64bit integer)
664 Depth of the tree used for fast key lookup.
665 JournalBytes (unsigned 64bit integer)
667 Total number of bytes written to the journal since startup.
668 JournalRotate (unsigned 64bit integer)
670 Number of times the journal has been rotated since startup.
671
673 SIGINT and SIGTERM
674 The daemon exits normally on receipt of either of these signals.
675 Pending updates are handled in accordance with the -j and -F
676 options.
677 SIGUSR1
679 The daemon exits AFTER flushing all updates out to disk. This may
680 take a while.
681 SIGUSR2
683 The daemon exits immediately, without flushing updates out to disk.
684 Pending updates will be replayed from the journal when the daemon
685 starts up again. WARNING: if journaling (-j) is NOT enabled, any
686 pending updates WILL BE LOST.
687
689 No known bugs at the moment.
690
692 rrdtool, rrdgraph
693
695 Florian Forster <octo at verplant.org>
696 Both rrdcached and this manual page have been written by Florian.
698
700 kevin brintnall <kbrint@rufus.net> Steve Shipway
701 <steve@steveshipway.org> Martin Sperl <rrdtool@martin.sperl.org>
7021.6.999 2017-03-10 RRDCACHED(1)