< inn.conf | Russ Allbery > Software > INN > INN CURRENT Documentation | moderators > |
(Configuration file for innfeed)
The configuration file innfeed.conf in pathetc is
used to control the innfeed(8) program. It is a fairly free-format file
that consists of three types of entries: key:value,
peer and group. Comments are from the hash character
#
to the end of the line.
key:value entries are a keyword and a value separated by a colon (which can itself be surrounded by whitespace). For example:
max-connections: 10
A legal key starts with a letter and contains only letters,
digits, and the _
and -
characters. There are 5 different
types of values: integers, floating-point numbers, characters, booleans,
and strings.
Integer and floating-point numbers are as to be expected, except that
exponents in floating-point numbers are not supported. A boolean value
is either true
or false
(case is not significant). A
character value is a single-quoted character as defined by the C-language.
A string value is any other sequence of characters. If the string needs
to contain whitespace, then it must be quoted with double quotes, and uses
the same format for embedding non-printing characters as normal C-language
string.
Peer entries look like:
peer <name> { # body ... }
The word peer
is required. The <name> is the same as the
site name in INN's newsfeeds configuration file. The body
of a peer entry contains some number (possibly zero) of
key:value entries.
Group entries look like:
group <name> { # body ... }
The word group
is required. The <name> is any string
valid as a key. The body of a group entry contains any number of the
three types of entries. So key:value pairs can be
defined inside a group, and peers can be nested inside a group, and other
groups can be nested inside a group.
key:value entries that are defined outside of all peer and group entries are said to be at "global scope". There are global key:value entries that apply to the process as a whole (for example the location of the backlog file directory), and there are global key:value entries that act as defaults for peers. When innfeed looks for a specific value in a peer entry (for example, the maximum number of connections to set up), if the value is not defined in the peer entry, then the enclosing groups are examined for the entry (starting at the closest enclosing group). If there are no enclosing groups, or the enclosing groups do not define the key:value, then the value at global scope is used.
A small example could be:
# Global value applied to all peers that have # no value of their own. max-connections: 5 # A peer definition. "uunet" is the name used by innd # in the newsfeeds configuration file. peer uunet { ip-name: usenet1.uu.net } peer vixie { ip-name: gw.home.vix.com max-connections: 10 # Override global value. } # A group of two peers which can handle more connections # than normal. group fast-sites { max-connections: 15 # Another peer. The "max-connections" value from the # "fast-sites" group scope is used. The "ip-name" value # defaults to the peer's name. peer data.ramona.vix.com { } peer bb.home.vix.com { max-connections: 20 # He can really cook. } }
Given the above configuration file, the defined peers would have the following values for the max-connections key:
uunet 5 vixie 10 data.ramona.vix.com 15 bb.home.vix.com 20
innfeed ignores key:value pairs it is not interested in. Some configuration file values can be set via a command-line option, in which case that setting overrides the settings in the file.
Configuration files can be included in other configuration files via the syntax:
$INCLUDE filename
There is a maximum nesting depth of 10.
For a fuller example configuration file, see the supplied innfeed.conf.
The following listing show all the keys that apply to the process as whole. These are not required (compiled-in defaults are used where needed).
This key requires a pathname value and defaults to patharticles in inn.conf. It specifies where the top of the article spool is. This corresponds to the -a command-line option.
This key requires a pathname value. It specifies the pathname (relative to the backlog-directory value) that should be read in funnel-file mode. This corresponds to giving a filename as an argument on the command-line (i.e. its presence also implies that funnel-file mode should be used).
The default is unset; innfeed then runs in channel or batch mode.
This key requires a pathname value and defaults to innfeed.pid. It specifies the pathname (relative to pathrun in inn.conf) where the pid of the innfeed process should be stored. This corresponds to the -p command-line option.
This key defines the debug level for the process. Default is 0
. A
non-zero number generates a lot of messages to stderr, or to the
config-defined log-file. This corresponds to the -d
command-line option.
If a file named innfeed.debug exists in the
pathlog directory (as set in inn.conf), then
debug-level is automatically set to 1
. This is a cheap
way of avoiding continual reloading of the newsfeeds file
when debugging. Note that debug messages still go to log-file.
This key requires a boolean value and defaults to false (the debug file is allowed to grow without bound). If set to true, this file is truncated when its size reaches a certain limit. See backlog-limit for more details.
This key requires a positive integer. The default value is 2
. It
defines the number of seconds to wait when innfeed (or a fork)
starts, before beginning to open connections to remote hosts.
This key requires a boolean value and defaults to false. If set to true, when innfeed receives a SIGTERM or SIGQUIT signal, it will close its listeners as soon as it can, even if it means dropping articles.
This key requires a boolean value and defaults to true. When
innfeed is given file names to send (a fairly rare use case)
instead of storage API tokens, it specifies whether mmaping should be used
if innfeed has been built with mmap(2) support. If article data on
disk is not in NNTP-ready format (CR/LF at the end of each line), then
after mmaping, the article is read into memory and fixed up, so mmaping
has no positive effect (and possibly some negative effect depending on
your system), and so in such a case this value should be false
,
which corresponds to the -M command-line option.
This key requires a pathname value and defaults to innfeed.log. It specifies where any logging messages that could not be sent via syslog(3) should go (such as those generated when a positive value for debug-value is used). This corresponds to the -l command-line option.
This pathname is relative to pathlog in inn.conf.
This key requires a format string suitable for strftime(3). It is used
for messages sent via syslog(3) and to the status-file. Default
value is %a %b %d %H:%M:%S %Y
.
This key requires a pathname value and defaults to innfeed. It specifies where the current innfeed process should store backlog files. This corresponds to the -b command-line option.
This pathname is relative to pathspool in inn.conf.
This key requires a positive integer value and defaults to 5
. It
specifies how many articles should be kept on the backlog file queue
before starting to write new entries to disk.
This key requires a positive integer value and defaults to 30
. It
specifies how many seconds elapse between checkpoints of the input backlog
file. Too small a number will mean frequent disk accesses; too large a
number will mean after a crash, innfeed will re-offer more
already-processed articles than necessary.
This key requires a positive integer value and defaults to 600
. It
specifies how many seconds elapse before each check for externally
generated backlog files that are to be picked up and processed.
This key requires a positive integer value and defaults to 60
. It
specifies how many seconds elapse before innfeed checks for a
manually created backlog file and moves the output backlog file to the
input backlog file.
This key requires a positive integer value and defaults to 900
. It
defines the number of seconds between attempts to re-lookup host
information that previously failed to be resolved.
This key requires a positive integer value and defaults to 86400
.
It defines the number of seconds between refreshes of name to address DNS
translation. This is so long-running processes do not get stuck with
stale data, should peer IP addresses change.
This key requires a boolean value and defaults to false. It specifies whether the status-file should be HTML-ified.
This key requires a pathname value and defaults to innfeed.status. An absolute pathname can be used. It specifies the pathname (relative to pathhttp when gen-html is true; otherwise, pathlog as set in inn.conf) where the periodic status of the innfeed process should be stored. This corresponds to the -S command-line option.
This key requires a boolean value and defaults to false. If the value is true, then whenever the transmission statistics for a peer are logged, each active connection logs its own statistics. This corresponds to the -z command-line option.
This key requires a positive integer value and defaults to 10
. It
defines how many articles will be held internally for a peer before new
arrivals cause article information to be spooled to the backlog file.
This key requires a positive integer value and defaults to 600
. It
defines how many seconds innfeed waits between generating
statistics on transfer rates.
This key requires a positive integer value and defaults to 43200
.
It defines how many seconds innfeed waits before resetting all
internal transfer counters back to zero (after logging one final time).
This is so a innfeed process running more than a day will generate
"final" stats that will be picked up by logfile processing scripts.
This key requires a positive integer value and defaults to 30
. It
defines how many seconds to first wait before retrying to reconnect after
a connection failure. If the next attempt fails too, then the reconnect
time is approximately doubled until the connection succeeds, or
max-reconnect-time is reached.
This key requires an integer value and defaults to 3600
. It
defines the maximum number of seconds to wait between attempt to reconnect
to a peer. The initial value for reconnection attempts is defined by
initial-reconnect-time, and it is doubled after each failure, up
to this value.
This key requires a non-negative integer value and defaults to 0
.
If the value is greater than zero, then whenever a network socket file
descriptor is created and it has a value less than this, the file
descriptor will be dup'ed to bring the value up greater than this. This
is to leave lower numbered file descriptors free for stdio. Certain
systems, Sun's in particular, require this. SunOS 4.1.x usually requires
a value of 128 and Solaris requires a value of 256. The default if this
is not specified, is 0
.
The following keys are used with imapfeed to authenticate to a remote host. Several parameters may be included at global scope:
The authname is who you want to authenticate as.
This is the appropriate password for authname.
The username is who you want to "act" as, that is, who is actually going to be using the server.
In this case, the "realm" is the realm in which the specified authname is valid. Currently this is only needed by the DIGEST-MD5 SASL mechanism.
A printf(3)-style format string for creating the envelope recipient
address. The pattern MUST include a single string specifier which will be
replaced with the newgroup (e.g. bb+%s
). The default is
+%s
.
An optional printf(3)-style format string for creating a To: header field
to be prepended to the article. The pattern MUST include a single string
specifier which will be replaced with the newgroup (e.g.
post+%s@domain
). If not specified, the To: header field will not
be prepended.
All the key:value pairs mentioned in this section can be specified at global scope. They may also be specified inside a group or peer definition. Note that when peers are added dynamically (i.e. when innfeed receives an article for an unspecified peer), it will add the peer site using the parameters specified at global scope.
No keys are currently required. They all have a default value, if not present in the configuration file.
The following keys are optional:
This key requires a non-negative integer value. The default value is
600
. If no articles need to be sent to the peer for this many
seconds, then the peer is considered idle and all its active connections
are torn down.
This key requires a non-negative integer value. The default value is
300
. It defines the maximum amount of time to wait for a response
from the peer after issuing a command.
This key requires a non-negative integer value. The default value is
1
. It defines the number of connections to be opened immediately
when setting up a peer binding. A value of 0
means no connections
will be created until an article needs to be sent.
This key requires a positive integer value. The default value is 2
but may be increased if needed or for large feeds. It defines the maximum
number of connections to run in parallel to the peer. A value of 0
specifies an unlimited number of maximum connections. In general, use of
an unlimited number of maximum connections is not recommended. Do not
ever set max-connections to zero with dynamic-method 0
set, as this will saturate peer hosts with connections.
This key requires a positive integer value and defaults to 86400
.
It is the maximum number of seconds a connection should be kept open.
Some NNTP servers do not deal well with connections being held open for
long periods.
This key requires an integer value between 0 and 3. The default value is
3
. It controls how connections are opened, up to the maximum
specified by max-connections. In general (and specifically, with
dynamic-method 0), a new connection is opened when the current
number of connections is below max-connections, and an article is
to be sent while no current connections are idle. Without further
restraint (i.e. using dynamic-method 0), in practice this means
that max-connections connections are established while articles
are being sent. Use of other dynamic-method settings imposes a
further limit on the amount of connections opened below that specified by
max-connections. This limit is calculated in different ways,
depending of the value of dynamic-method.
Users should note that adding additional connections is not always productive -- just because opening twice as many connections results in a small percentage increase of articles accepted by the remote peer, this may be at considerable resource cost both locally and at the remote site, whereas the remote site might well have received the extra articles sent from another peer a fraction of a second later. Opening large numbers of connections is considered antisocial.
The meanings of the various settings are:
Increase of connections up to max-connections is unrestrained.
Connections are increased (up to max-connections) and decreased so as to maximize the number of articles per second sent, while using the fewest connections to do this.
Connections are increased (up to max-connections) and decreased so as to keep the queue of articles to be sent within the bounds set by dynamic-backlog-low and dynamic-backlog-high, while using the minimum resources possible. As the queue will tend to fill if the site is not keeping up, this method ensures that the maximum number of articles are offered to the peer while using the minimum number of connections to achieve this.
This method uses a combination of methods 1 and 2 above. For sites accepting a large percentage of articles, method 2 will be used to ensure these sites are offered as complete a feed as possible. For sites accepting a small percentage of articles, method 1 is used, to minimize remote resource usage. For intermediate sites, an appropriate combination is used.
This key requires a floating-point value between 0 and 100. It represents
(as a percentage) the low water mark for the host queue. If the host
queue falls below this level while using dynamic-method 2 or 3,
and if 2 or more connections are open, innfeed will attempt to drop
connections to the host. An Infinite Impulse Response (IIR) filter is
applied to the value to prevent connection flap (see
dynamic-filter). The default value is 20.0
. This value
must be smaller than dynamic-backlog-high.
This key requires a floating-point value between 0 and 100. It represents
(as a percentage) the high water mark for the host queue. If the host
queue rises above this level while using dynamic-method 2 or 3,
and if less than max-connections are open to the host,
innfeed will attempt to open further connections to the host. An
Infinite Impulse Response (IIR) filter is applied to the value to prevent
connection flap (see dynamic-filter). The default value is
50.0
. This value must be larger than dynamic-backlog-low.
This key requires a floating-point value between 0 and 1. It represents
the filter coefficient used by the Infinite Impulse Response (IIR) filter
used to implement dynamic-method 2 and 3. The default value of
this filter is 0.7
, giving a time constant of 1/(1-0.7) articles.
Higher values will result in slower response to queue fullness changes;
lower values in faster response.
This key requires a positive integer value. The default value is
20
. It defines the maximum number of articles to process at one
time when using streaming to transmit to a peer. Larger numbers mean more
memory consumed as articles usually get pulled into memory (see the
description of use-mmap).
This key requires a boolean value. Its default value is true. It defines whether streaming commands are used to transmit articles to the peers.
This key requires a floating-point number which must be in the range
[0.0, 100.0]. When running transmitting with the
streaming commands, innfeed attempts an optimization called
"no-CHECK mode". This involves not asking the peer if it wants
the article, but just sending it. This optimization occurs when the
percentage of the articles the peer has accepted gets larger than this
number. If this value is set to 100.0
, then this effectively turns
off no-CHECK mode, as the percentage can never get above 100.0. If this
value is too small, then the number of articles the peer rejects will get
bigger (and your bandwidth will be wasted). The default value of
95.0
usually works pretty well.
This key requires a floating-point number which must be in the range
[0.0, 100.0], and it must be smaller that the value
for no-check-high. When running in no-CHECK mode, as described
above, if the percentage of articles the remote server accepts drops below
this number, then the no-CHECK optimization is turned off until the
percentage gets above the no-check-high value again. If there is
small difference between this and the no-check-high value (less
than about 5.0), then innfeed may frequently go in and out of
no-CHECK mode. If the difference is too big, then it will make it harder
to get out of no-CHECK mode when necessary (wasting bandwidth). Keeping
this to between 5.0 and 10.0 less than no-check-high usually works
pretty well. The default value is 90.0
.
This is a floating-point value representing the time constant, in
articles, over which the CHECK/no-CHECK calculations are done. The
default value is 50.0
, which will implement an Infinite Impulse
Response (IIR) filter of time constant 50. This roughly equates to making
a decision about the mode over the previous 50 articles. A higher number
will result in a slower response to changing percentages of articles
accepted; a lower number will result in a faster response.
This key requires a positive integer value. It defines the TCP/IP port number to use when connecting to the remote. Usually, port number 119 is used, which is the default value.
This key requires a boolean value. By default, it is set to false.
Setting it to true is the same as setting bindaddress6 to
none
and removing bindaddress from none
if it was
set.
This key requires a boolean value. By default, it is set to false. When set to true, and a peer replies with code 431 or 436 (try again later), innfeed just drops the article and does not try to re-send it. This is useful for some peers that keep on deferring articles for a long time to prevent innfeed from trying to offer the same article over and over again.
This key requires a boolean value. By default, it is set to false. When
set to true, innfeed will attempt to use a connection with the
least queue size (or the first empty connection). If this key is set to
true, it is recommended that dynamic-method be set to 0
.
This allows for article propagation with the least delay.
This key requires a boolean value. It specifies whether spooling should be enabled (false, the default) or disabled (true). Note that when no-backlog is set, articles reported as spooled are actually silently discarded.
This key requires a non-negative integer value. If the number is 0
(the default), then backlog files are allowed to grow without bound when
the peer is unable to keep up with the article flow. If this number is
greater than 0, then it specifies the size (in bytes) the backlog file
should get truncated to when the backlog file reaches a certain limit.
The limit depends on whether backlog-factor or
backlog-limit-highwater is used.
This parameter also applies to the debug file when debug-shrinking is set to true, and has the same effect on this file as the one has on backlog files.
This key requires a floating-point value, which must be larger than 1.0. It is used in conjunction with the peer key backlog-limit. If backlog-limit has a value greater than zero, then when the backlog file gets larger than the value backlog-limit * backlog-factor, then the backlog file will be truncated to the size backlog-limit.
For example, if backlog-limit has a value of 1000000
, and
backlog-factor has a value of 2.0
, then when the backlog
file gets to be larger than 2000000 bytes in size, it will be truncated to
1000000 bytes. The front portion of the file is removed, and the trimming
happens on line boundaries, so the final size may be a bit less than this
number. If backlog-limit-highwater is defined too, then
backlog-factor takes precedence. The default value of
backlog-factor is 1.1
.
This parameter also applies to the debug file when debug-shrinking is set to true, and has the same effect on this file as the one has on backlog files.
This key requires a positive integer value that must be larger than the
value for backlog-limit. The default value is 0
.
If the size of the backlog file gets larger than this value (in bytes), then the backlog file will be shrunk down to the size of backlog-limit. If both backlog-factor and backlog-limit-highwater are defined, then the value of backlog-factor is used.
This parameter also applies to the debug file when debug-shrinking is set to true, and has the same effect on this file as the one has on backlog files.
This key requires a boolean value. By default it is set to false. When set to true, the backlog is fed before new files. This is intended to enforce in-order delivery, so setting this to true when initial-connections or max-connections is more than 1 is inconsistent.
This key requires a string value. It specifies which outgoing IPv4
address innfeed should bind the local end of its connection to. It
must be an IPv4 address in dotted-quad format (nnn.nnn.nnn.nnn),
any
, or none
. If not set or set to any
,
innfeed defaults to letting the kernel choose this address. If set
to none
, innfeed will not use IPv4 for outgoing connections
to peers in this scope (i.e. it forces IPv6).
If not set in innfeed.conf, innfeed defaults to the value of sourceaddress from inn.conf (which by default is unset).
This key requires a string value. It behaves like bindaddress
except for outgoing IPv6 connections. It must be in numeric IPv6 format
(note that a value containing colons must be enclosed in double quotes),
any
, or none
. If set to none
, innfeed will
not use IPv6 for outgoing connections to peers in this scope.
If not set in innfeed.conf, innfeed defaults to the value of sourceaddress6 from inn.conf (which by default is unset).
This key requires a string value. If the value is defined, then innfeed tries to authenticate by AUTHINFO USER and this value used for user name. password must also be defined, if this key is defined.
This key requires a string value. The value is the password used for AUTHINFO PASS. username must also be defined, if this key is defined.
As previously explained, the peer definitions can contain redefinitions of any of the key:value pairs described in the section about global peer defaults above. There is one key:value pair that is specific to a peer definition.
This key requires a word value. The word is either one of the host's FQDNs, or the dotted-quad IP address of the peer for IPv4, or the colon-separated IP address of the peer for IPv6. If this value is not specified, then the name of the peer in the enclosing peer block is taken to also be its ip-name.
If innfeed gets a SIGHUP signal, then it will reread the configuration file. All values at global scope except for backlog-directory can be changed (although note that bindaddress and bindaddress6 changes will only affect new connections).
Any new peers are added and any missing peers have their connections closed.
The log file is also reopened.
For a comprehensive example, see the sample innfeed.conf distributed with INN and installed as a starting point.
Here are examples of how to format values:
eg-string: "New\tconfig\tfile\n" eg-long-string: "A long string that goes over multiple lines. The newline is kept in the string except when quoted with a backslash \ as here." eg-simple-string: A-no-quote-string eg-integer: 10 eg-boolean: true eg-char: 'a' eg-ctrl-g: '\007'
Written by James Brister <brister@vix.com> for InterNetNews. Converted to POD by Julien Elie.
Earlier versions of innfeed (up to 0.10.1) were shipped separately; innfeed is now part of INN and shares the same version number. Please note that the innfeed.conf format has changed dramatically since version 0.9.3.
$Id: innfeed.conf.pod 10178 2017-09-18 20:13:08Z iulius $
inn.conf(5), innfeed(8), newsfeeds(5).
< inn.conf | Russ Allbery > Software > INN > INN CURRENT Documentation | moderators > |