3 innduct \- quickly and reliably stream Usenet articles to remote site
11 implements NNTP peer-to-peer news transmission including the streaming
12 extensions, for sending news articles to a remote site.
14 You need to run one instance of innduct for each peer site. innduct
15 manages its interaction with innd, including flushing the feed as
16 appropriate, etc., so that articles are transmitted quickly, and
17 manages the retransmission of its own backlog. innduct includes the
18 locking necessary to avoid multiple simutaneous invocations.
20 By default, innduct reads the default feedfile corresponding to
24 .IR pathoutgoing / site )
25 and feeds it via NNTP, streaming if possible, to the host
30 is not specified, it defaults to
33 innduct daemonises after argument parsing, and all logging (including
34 error messages) are sent to syslog (facility
37 The best way to run innduct is probably to periodically invoke innduct
38 for each feed (e.g. from cron), passing innduct it the
40 option to arrange that it silently exits if an innduct is already
41 running for that site.
42 .SH INNDUCT VS INNFEED/NNTPSEND/INNXMIT
45 does roughly the same thing as innduct. However, the way it receives
46 information from innd can result in articles being lost (not offered
47 to peers) if innfeed crashes for any reason. This is an inherent
48 defect in the innd channel feed protocol. innduct uses a file feed,
49 constantly "tailing" the feed file, and where implemented uses
51 to reduce the latency which would come from having to constantly poll
52 the feed file. innfeed is capable of feeding multiple peers from a
53 single innfeed instance, whereas each innduct process handles exactly
54 one peer. innduct is much smaller and simpler, at <4kloc to innfeed's
55 25kloc. innfeed needs a separate wrapper script or similar
56 infrastructure (of which there is an example in its manpage), whereas
57 innduct can be run directly and doesn't need help from shell scripts.
60 processes feed files in batch mode. That is, you have to periodically
61 invoke nntpsend, and when you do, the feed is flushed and articles
62 which arrived before the flush are sent to the peer. This introduces
63 a batching delay, and also means that the NNTP connection to the peer
64 needs to be remade at each batch. nntpsend (which uses innxmit)
65 cannot make use of multiple connections to a single peer site.
66 However, nntpsend can be left to find automatically which sites need
71 is the actual NNTP feeder program used by nntpsend.
74 .BR \-f | \-\-feedfile= \fIfeedfile\fR
77 If the specified value ends in a
79 it is taken as a directory to use as if it were
81 and the actual feed file used is
82 .IR specified_feedfile / site .
84 .BR \-q | \-\-quiet-multiple
85 Makes innduct silently exit (with status 0) if another innduct holds
86 the lock for the site. Without \fB-q\fR, this causes a fatal error to
87 be logged and a nonzero exit.
90 Do not daemonise. innduct runs in the foreground and all messages
91 (including all debug messages) are written to stderr.
94 Do not try to use the streaming extensions to NNTP (for use eg if the
95 peer can't cope when we send MODE STREAM).
98 Do not try to use the file change monitoring support to watch for
99 writes by innd to the feed file; poll it instead. (If file monitoring
100 is not compiled in, this option just downgrades the log message which
101 warns about this situation.)
103 .BR \-C | \-\-inndconf= \fIFILE\fR
106 instead of the default
109 .BR \-\-cli= \fIDIR\fR / |\fIPATH\fR
110 Bind the control command line socket to
112 (if the value ends with a
116 (if it doesn't). See CONTROLLING INNDUCT, below.
117 This option may be essential, if the
120 is too long, as there is a fairly short limit on the paths to AF_UNIX
121 sockets. The default is
122 .IR feedfile \fB_cli\fR.
127 at the remote site rather than to the NNTP port (119).
130 Just print a brief usage message and list of the options to stdout.
132 You should not normally need to adjust these. Time intervals may
133 specified in seconds, or as a number followed by one of the following
136 .BR "sec min hour day" ,
139 .BI \-\-max-connections= max
140 Restricts the maximum number of simultaneous NNTP connections used by
145 There is no global limit on the number of connections.
147 .BI \-\-max-queue-per-conn= per-conn-max
148 Restricts the maximum number of outstanding articles queued on any
149 particular connection to
151 (Non-streaming connections can only handle one article at a time.)
155 .BI \-\-max-queue-per-file= max
156 Restricts the maximum number articles read into core from any one
159 The default is twice the value of per-conn-max.
161 .BI \-\-feedfile-flush-size= bytes
162 Specifies that innduct should flush the feed and start a new feedfile
163 when the existing feedfile size exceeds
165 the effect is that the innduct will try to avoid the various
166 batchfiles growing much beyond this size while the link to the peer is
167 working. The default is
170 .BI \-\-period-interval= PERIOD-INTERVAL
171 Specifies wakup interval and period granularity.
172 innduct wakes up every
174 to do various housekeeping checks. Also, many of the timeout and
175 rescan intervals (those specified in this manual as
177 are rounded up to the next multiple of
178 .IR PERIOD-INTERVAL .
180 .BI \-\-connection-timeout= TIME
181 How long to allow for a connection setup attempt before giving up.
185 .BI \-\-stuck-flush-timeout= TIME
186 How long to wait for innd to respond to a flush request before giving
190 .BI \-\-feedfile-poll= TIME
191 How often to poll the feedfile for new articles written by innd
194 or equivalent) is not available. (When file monitoring is available,
195 there is no need for periodic checks and we wake immediately up
196 whenever the feedfile changes.)
200 .BI \-\-no-check-proportion= PERCENT
201 If the moving average of the proportion of articles being accepted
202 (rather than declined) by the peer exceeds this value, innduct uses
203 "no check mode" - ie it just sends the peer the articles with TAKETHIS
204 rather than checking first with CHECK whether the article is wanted.
205 This only affects streaming connections. The default is
209 .BI \-\-no-check-response-time= ARTICLES
210 The moving average mentioned above is an alpha-smoothed value with a
216 .BI \-\-reconnect-interval= RECONNECT-PERIOD
217 Limits initiation of new connections to one each
218 .IR RECONNECT-PERIOD .
219 This applies to reconnections if the peer has been down, and also to
220 ramping up the number of connections we are using after startup or in
221 response to an article flood. The default is
224 .BI \-\-flush-retry-interval= PERIOD
225 If our attempt to flush the feed failed (usually this will be because
226 innd is not running), try again after
231 .BI \-\-earliest-deferred-retry= PERIOD
232 When the peer responds to our offer of an article with a 431 or 436
233 NNTP response code, indicating that the article has already been
234 offered to it by another of its peers, and that we should try again,
237 before offering the article again. The default is
240 .BI \-\-backlog-rescan-interval= BACKLOG-SCAN-PERIOD
241 We scan the directory containing
243 for backlog files at least every
244 .IR BACKLOG-SCAN-PERIOD ,
245 in case the administrator has manually dropped in a file there for
249 .BI \-\-max-flush-interval= PERIOD
250 We flush the feed at least every
252 even if the current instance of the feedfile has not reached the size
257 .BI \-\-max-flush-interval= PERIOD
258 We flush the feed and start a new feedfile at least every
260 even if the current instance of the feedfile has not reached the size
265 .BI \-\-flush-finish-timeout= FLUSH-FINISH-PERIOD
267 .IR FLUSH-FINISH-PERIOD
268 ago, and are still trying to finish processing articles that were
269 written to the old feed file, we forcibly and violently make sure that
270 we do by abandoning and deferring all the work (which includes
271 unceremoniously dropping any connections on which we've sent some of
272 those articles but not yet had replies, as they're probably stuck
277 .BI \-\-idle-timeout= PERIOD
278 Connections which have had no activity for
280 will be closed. This includes connections where we have sent commands
281 or articles but have not yet had the responses, so this same value
282 doubles as the timeout after which we conclude that the peer is
283 unresponsive or the connection has become broken.
287 .BI \-\-max-bad-input-data-ratio= PERCENT
288 We tolerate up to this proportion of badly-formatted lines in the
289 feedfile and other input files. Every badly-formatted line is logged,
290 but if there are too many we conclude that the corruption to our
291 on-disk data is too severe, and crash; to successfully restart,
292 administrator intervention will be required. This avoids flooding the
293 logs with warnings and also arranges to abort earlyish if an attempt
294 is made to process a file in the wrong format.
299 .BI \-\-max-bad-input-data-init= LINES
300 Additionally, we tolerate this number of additional badly-formatted
301 lines, so that if the badly-formatted lines are a few but at the start
302 of the file, we don't crash immediately.
305 (which would suffice to ignore one whole corrupt 4096-byte disk block
306 filled with random data, or one corrupt 1024-byte disk block filled
307 with an inappropriate text file with a mean line length of at least
309 .SH CONTROLLING INNDUCT
310 If you tell innd to drop the feed, innduct will (when it notices,
311 which will normally be the next time it decides flushes) finish up the
312 articles it has in hand now, and then exit. It is harmless to cause
313 innd to flush the feed (but innduct won't notice and flushing won't
314 start a new feedfile; you have to leave that to innduct).
316 If you want to stop innduct you can send it SIGTERM or SIGINT, or the
318 control command, in which case it will report statistics so far and
319 quickly exit. If innduct receives SIGKILL nothing will be broken or
320 corrupted; you just won't see some of the article stats.
322 innduct listens on an AF_UNIX socket, and provides a command-line
323 interface which can be used to trigger various events and for
324 debugging. The socket is called
326 and when connected reads and writes lines (with unix line endings).
327 It can most easily be accessed with a program like
331 .IR feedfile \fB_cli\fR)
337 The following control commands are supported:
340 Print a list of all the commands understood. This list includes
341 undocumented commands which mess with innduct's internal state and
342 should only be used by a developer in conjuction with the innduct
346 Start a new feed file and trigger a flush of the feed. (Or, cause
348 .I FLUSH-FINISH-PERIOD
349 to expire early, forcibly completing a previously started flush.)
352 Log statistics and exit. (Same effect as SIGTERM or SIGINT.)
355 Writes information about innduct's state to a plain text file
356 .IR feedfile \fB_dump\fR.
357 This overwrites any previous dump.
359 is a summary including general state and a list of connections;
361 also includes information about each article innduct is dealing with.
364 Requests that innduct rescan for new backlog files at the next
366 poll. Normally innduct assumes that any backlog files dropped in by
367 the administrator are not urgent and may not get around to noticing
369 .IR BACKLOG-SCAN-PERIOD .
372 Resets the connection startup delay counter so that innduct may
373 consider making a new connection to the peer right away, regardless
375 .IR RECONNECT-PERIOD .
376 A connection attempt will still only be made if innduct feels that it
377 needs one, and innduct may wait up to
379 before actually starting the attempt.
380 .IR BACKLOG-SCAN-PERIOD .
384 An instance of innduct is already running for this
391 The feed has been dropped by innd, and we (or previous innducts) have
392 successfully offered all the old articles to the peer site. Our work
396 innduct was invoked with bad options or command line arguments. The
397 error message will be printed to stderr, and also (if any options or
398 arguments were passed at all) to syslog with severity
402 Things are going wrong, hopefully shortage of memory, system file
403 table entries; disk IO problems; disk full; etc. The specifics of the
404 error will be logged to syslog with severity
406 (if syslog is working!)
409 Things are going badly wrong in an unexpected way: system calls which
410 are not expected to fail are doing so, or the protocol for
411 communicating with innd is being violated, or some such. Details will
412 be logged with severity
414 (if syslog is working!)
417 These exit statuses are used by children forked by innduct to
418 communicate to the parent. You should not see them. If you do, it is
421 innduct dances a somewhat complicated dance with innd to make sure
422 that everything goes smoothly and that there are no races. (See the
423 two ascii-art diagrams in innduct.c for details of the protocol.) Do
424 not mess with the feedfile and other associated files, other than as
427 .IP \fIpathoutgoing\fR/\fIsite\fR
428 .IX Item "default feedfile"
433 Main feed file as specified in
435 This and other batchfiles used by innduct contains lines each of which
437 \& \fItoken\fR \fImessageid\fR
438 where \fItoken\fR is the inn storage API token. Such lines can be
439 written by \fBTf,Wnm\fR in a \fInewsfeeds\fR(5) entry. During
440 processing, innduct overwrites lines in the batch files which
441 correspond to articles it has processed: the line is replaced with
442 one containing only spaces. Only innd should create this file, and
443 only innduct should remove it.
444 .IP \fIfeedfile\fR_lock
446 Lockfile, preventing multiple innduct invocations for the same
447 feed. A process holds this lock after it has opened the lockfile,
448 made an fcntl F_SETLK call, and then checked with stat and fstat that
449 the file it now has open and has locked still has the name
450 \fIfeedfile\fR_lock. (Only) the lockholder may delete the lockfile.
451 For your convenience, after the lockfile is locked,
458 are all written to the lockfile. NB that stale lockfiles may contain
459 stale data so this information should not be relied on other than for
461 .IP \fIfeedfile\fR_flushing
462 .IX Item "flushing file"
463 Batch file: the main feedfile is renamed to this filename by innduct
464 before it asks inn to flush the feed. Only innduct should create or
466 .IP \fIfeedfile\fR_defer
467 .IX Item "flushing file"
468 Batch file containing details of articles whose transmission has
469 recently been deferred at the request of the recipient site. Created,
470 written, read and removed by innduct.
471 .IP \fIfeedfile\fR_backlog.\fItime_t\fR.\fIinum\fR
472 .IX Item "backlog file"
473 Batch file containing details of articles whose transmission has less
474 recently been deferred at the request of the recipient site. Created
475 by innduct, and will also be read and removed by innduct. However you
476 (the administrator) may also safely remove backlog files.
477 .IP \fIfeedfile\fR_backlog\fIsomething\fR
478 .IX Item "manual backlog file"
479 Batch file manually provided by the administrator. The file should be
480 complete and ready to process at the time it is renamed or hardlinked
481 to this name. innduct will then automatically find and read and
482 process it and eventually remove it. The administrator may also
483 safely remove backlog files. \fIsomething\fR may not contain \fB#\fR
484 \fB~\fR or \fB/\fR. Be sure to have finished writing the file before
485 you rename it to match the pattern \fIfeedfile\fR\fB_backlog\fR*, as
486 otherwise innduct may find and process the file and read it to EOF
487 before you have finished creating it.
488 .IP \fIfeedfile\fR_cli
489 .IX Item "control command line socket"
490 Default AF_UNIX listening socket for the control command line. See
491 CONTROLLING INNDUCT, above.
492 .IP \fIfeedfile\fR_dump
493 .IX Item "debug dump file"
494 On request via a control connection innduct dumps a summary of its
495 state to this text file. This is mostly useful for debugging.
496 .IP /etc/news/inn.conf
500 if none is specified, for finding how to communicate with innd,
506 Written by Ian Jackson <ijackson@chiark.greenend.org.uk>
~ian / innduct.git / blob