mirmon - monitor the state of mirrors
mirmon [ -v ] [ -q ] [ -t timeout ] [ -get opt ] [ -c conf ]
option v : be verbose option q : be quiet option t : set timeout [ default 300 ] ; option get : 'all' : probe all sites : 'update' : probe a selection of the sites (see doc) option c : configuration file ; default list : ./mirmon.conf $HOME/.mirmon.conf /etc/mirmon.conf ------------------------------------------------------------------- Documentation : the program contains 'pod' style documentation. Extract the doc with 'pod2text mirmon' or 'pod2html mirmon OUT', etc. -------------------------------------------------------------------
The program is intended to be run by cron every hour. 42 * * * * perl /path/to/mirmon -q -get update It quietly probes a subset of the sites in a given list, writes the results in the 'state' file and generates a web page with the results. The subset contains the sites that are new, bad and/or not probed for a specified time. When no 'get' option is specified, the program just generates a new web page from the last known state. The program checks the mirrors by running a (user specified) program on a pipe. A (user specified) number of probes is run in parallel using nonblocking IO. When something can be read from the pipe, it switches the pipe to blocking IO and reads one line from the pipe. Then it flushes and closes the pipe. No attempt is made to kill the probe. The probe should return something that looks like "1043625600\n", that is, a timestamp followed by a newline. The exit status of the probe is ignored.
A config file can be specified with the -c option. If -c is not used, the program looks for a config file in -- ./mirmon.conf -- $HOME/.mirmon.conf -- /etc/mirmon.conf
A config file looks like this : +-------------------------------------------------- |# lines that start with '#' are comment |# blank lines are ignored too |# tabs are replaced by a space | |# the config entries are 'key' and 'value' pairs |# a 'key' begins in column 1 |# the 'value' is the rest of the line |somekey A_val B_val ... |otherkey X_val Y_val ... | |# indented lines are glued |# the next three lines mean 'somekey part1 part2 part3' |somekey part1 | part2 | part3 | |# lines starting with a '+' are concatenated |# the next three lines mean 'somekey part1part2part3' |somekey part1 |+ part2 |+ part3 | |# lines starting with a '.' are glued too |# don't use a '.' on a line by itself |# 'somekey' gets the value "part1\n part2\n part3" |somekey part1 |. part2 |. part3 +--------------------------------------------------
Specify a short plaintext name for the project. project_name Apache project_name CTAN
Specify an url pointing to the 'home' of the project. project_url http://www.apache.org/
Specify the file containing the mirrors to probe. Two formats are supported : -- plain : lines like us http://www.tux.org/ nl http://apache.cs.uu.nl/dist/ -- apache : lines like those in the apache mirrors.list ftp us ftp://ftp.tux.org/pub/net/apache/dist/ firstname.lastname@example.org http nl http://apache.cs.uu.nl/dist/ email@example.com Specify the required format with 'list_style' (see below). The default style is 'plain'. If the url part of a line doesn't end in a slash ('/'), mirmon adds a slash and issues a warning unless it is in quiet mode.
Specify where the html report page is written.
Specify the directory where the icons can be found.
Specify the program+args to probe the mirrors. Example: probe /sw/bin/wget -q -O - -T %TIMEOUT% -t 1 %URL%TIME Before the program is started, %TIMEOUT% and %URL% are substituted with the proper timeout and url values. Here it is assumed that each hour the root server writes a timestamp in /path/to/archive/TIME, for instance with a crontab entry like 42 * * * * perl -e 'printf "%s\n", time' > /path/to/archive/TIME Mirmon reads one line of output from the probe and interprets the first word on that line as a timestamp ; for example : 1043625600 1043625600 Mon Jan 27 00:00:00 2003 1043625600 www.apache.org Mon Jan 27 00:00:00 2003
Specify where the file containing the state is written. The program reads this file on startup and writes the file when mirrors are probed (-get is specified).
Specify the file containing the country codes; The file should contain lines like us - united states nl - netherlands The mirmon package contains a recent ISO list.
Optionally specify the number of parallel probes (default 25).
Optionally specify the timeout for the probes (default 300). After the last probe is started, the program waits for <timeout> + 10 seconds, cleans up and exits.
Optionally specify (the SRC of the IMG of) a logo to be placed top right on the page. project_logo /icons/apache.gif project_logo http://www.apache.org/icons/...
Optionally specify some HTML to be placed before </HEAD>. htm_head <link REL=StyleSheet HREF="/style.css" TYPE="text/css">
Optionally specify some HTML to be placed near the top of the page. The supplied text is placed between <P> and </P>. htm_top testing 1, 2, 3
Optionally specify HTML to be placed near the bottom of the page. htm_foot <HR> <A HREF="..."><IMG SRC="..." BORDER=0></A> <HR>
Optionally specify where the age histogram must be placed. The default is 'top'.
For 'min_poll' see next item. A <time spec> is a number followed by a unit 's' (seconds), or 'm' (minutes), or 'h' (hours), or 'd' (days). For example '3d' (three days) or '36h' (36 hours).
Optionally specify the maximum probe interval. When the program is called with option '-get update', all sites are probed which are : -- new : the site appears in the list, but there is no known state -- bad : the last probe of the site was unsuccessful -- old : the last probe was more than 'max_poll' ago. Sites are not probed if the last probe was less than 'min_poll' ago. So, if you specify min_poll 4h max_poll 12h the 'reachable' sites are probed twice daily and the 'unreachable' sites are probed at most six times a day. The default 'min_poll' is '1h' (1 hour). The default 'max_poll' is '4h' (4 hours).
Optionally specify how often the mirrors are required to make an update. The default 'min_sync' is '1d' (1 day).
Optionally specify the maximum allowable sync interval. Sites exceeding the limit will be considered 'old'. The default 'max_sync' is '2d' (2 days).
With a low probablility, mirmon probes mirrors that would otherwise not be probed. In the long run, this balances the number of mirror probes over the hourly mirmon runs. Specifically, if there are N mirrors in the list and some mirmon run would probe K sites, on average (N-K)/N extra sites will be probed. If you don't want this behaviour, use 'no_randomize'.
Optionally specify the format ('plain' or 'apache') of the mirror-list. See the description of 'mirror_list' above. The default list_style is 'plain'.
Optionally specify a substitute url for a site. When access to a site is restricted (in Australia, for instance), another (sometimes secret) url can be used to probe the site. The <site> of an url is the part between '://' and the first '/'.
Optionally specify an environment variable.
Optionally specify a file to include. The specified file is processed 'in situ'. After the specified file is read and processed, config processing is resumed in the file where the 'include' was encountered. The 'include' depth is unlimited. However, it is a fatal error to include a file twice under the same name.
When the config processor encounters the 'show' command, it dumps the content of the current config to standout, if option -v is specified. This is intented for debugging.
When the config processor encounters the 'exit' command, it terminates the program. This is intented for debugging.
The state file consists of lines; one line per site. Each line consists of white space separated fields. The seven fields are :
The url as given in the mirror list.
The age of the site, or 'undef' if no probe was ever successful.
The status of the last probe.
The timestamp of the last successful probe or 'undef' if the site was never successfully probed.
The probe history is a list of 's' (for success) and 'f' (for failure) characters indicating the result of the probe. New results are appended whenever the site is probed.
The state history consists of a timestamp, a '-' char, and a list of chars indicating a past status: 's' (fresh), 'b' (oldish), 'f' (old) or 'z' (bad). The timestamp indicates when the state history was last updated. The state history is updated when the state file is updated and the last update of the history state was 24 (or more) hours ago. The status is determined by the site's age and a few configuration parameters. The details are explained in the legend of the report page.
The timestamp of the last probe.
The '#!' path for perl is probably wrong.