SYNOPSIS
pflogsumm.pl -[eq] [-d <today|yesterday>] [--detail <cnt>]
[--bounce_detail <cnt>] [--deferral_detail <cnt>]
[-h <cnt>] [-i|--ignore_case] [--iso_date_time] [--mailq]
[-m|--uucp_mung] [--no_bounce_detail] [--no_deferral_detail]
[--no_no_msg_size] [--no_reject_detail] [--no_smtpd_warnings]
[--problems_first] [--rej_add_from] [--reject_detail <cnt>]
[--smtp_detail <cnt>] [--smtpd_stats]
[--smtpd_warning_detail <cnt>] [--syslog_name=string]
[-u <cnt>] [--verbose_msg_detail] [--verp_mung[=<n>]]
[--zero_fill] [file1 [filen]]
pflogsumm.pl -[help|version]
If no file(s) specified, reads from stdin. Output is to stdout.
DESCRIPTION
Pflogsumm is a log analyzer/summarizer for the Postfix MTA. It is
designed to provide an over-view of Postfix activity, with just enough
detail to give the administrator a "heads up" for potential trouble
spots.
Pflogsumm generates summaries and, in some cases, detailed reports of
mail server traffic volumes, rejected and bounced email, and server
warnings, errors and panics.
OPTIONS
--bounce_detail <cnt>
Limit detailed bounce reports to the top <cnt>. 0
to suppress entirely.
-d today generate report for just today
-d yesterday generate report for just "yesterday"
--deferral_detail <cnt>
Limit detailed deferral reports to the top <cnt>. 0
to suppress entirely.
--detail <cnt>
Sets all --*_detail, -h and -u to <cnt>. Is
over-ridden by individual settings. --detail 0
suppresses *all* detail.
-e extended (extreme? excessive?) detail
Emit detailed reports. At present, this includes
only a per-message report, sorted by sender domain,
then user-in-domain, then by queue i.d.
WARNING: the data built to generate this report can
quickly consume very large amounts of memory if a
lot of log entries are processed!
-h <cnt> top <cnt> to display in host/domain reports.
0 = none.
See also: "-u" and "--*_detail" options for further
report-limiting options.
--help Emit short usage message and bail out.
(By happy coincidence, "-h" alone does much the same,
being as it requires a numeric argument :-). Yeah, I
know: lame.)
-i
--ignore_case Handle complete email address in a case-insensitive
manner.
Normally pflogsumm lower-cases only the host and
domain parts, leaving the user part alone. This
option causes the entire email address to be lower-
cased.
--iso_date_time
For summaries that contain date or time information,
use ISO 8601 standard formats (CCYY-MM-DD and HH:MM),
rather than "Mon DD CCYY" and "HHMM".
-m modify (mung?) UUCP-style bang-paths
--uucp_mung
This is for use when you have a mix of Internet-style
domain addresses and UUCP-style bang-paths in the log.
Upstream UUCP feeds sometimes mung Internet domain
style address into bang-paths. This option can
sometimes undo the "damage". For example:
"somehost.dom!username@foo" (where "foo" is the next
host upstream and "somehost.dom" was whence the email
originated) will get converted to
"[email protected]". This also affects the
extended detail report (-e), to help ensure that by-
domain-by-name sorting is more accurate.
--mailq Run "mailq" command at end of report.
Merely a convenience feature. (Assumes that "mailq"
is in $PATH. See "$mailqCmd" variable to path thisi
if desired.)
--no_bounce_detail
--no_deferral_detail
--no_reject_detail
These switches are depreciated in favour of
--bounce_detail, --deferral_detail and
--reject_detail, respectively.
Suppresses the printing of the following detailed
reports, respectively:
message bounce detail (by relay)
message deferral detail
message reject detail
See also: "-u" and "-h" for further report-limiting
options.
--no_no_msg_size
Do not emit report on "Messages with no size data".
Message size is reported only by the queue manager.
The message may be delivered long-enough after the
(last) qmgr log entry that the information is not in
the log(s) processed by a particular run of
pflogsumm.pl. This throws off "Recipients by message
size" and the total for "bytes delivered." These are
normally reported by pflogsumm as "Messages with no
size data."
--no_smtpd_warnings
This switch is depreciated in favour of
smtpd_warning_detail
On a busy mail server, say at an ISP, SMTPD warnings
can result in a rather sizeable report. This option
turns reporting them off.
--problems_first
Emit "problems" reports (bounces, defers, warnings,
etc.) before "normal" stats.
--rej_add_from
For those reject reports that list IP addresses or
host/domain names: append the email from address to
each listing. (Does not apply to "Improper use of
SMTP command pipelining" report.)
-q quiet - don't print headings for empty reports
note: headings for warning, fatal, and "master"
messages will always be printed.
--reject_detail <cnt>
Limit detailed smtpd reject, warn, hold and discard
reports to the top <cnt>. 0 to suppress entirely.
--smtp_detail <cnt>
Limit detailed smtp delivery reports to the top <cnt>.
0 to suppress entirely.
--smtpd_stats
Generate smtpd connection statistics.
The "per-day" report is not generated for single-day
reports. For multiple-day reports: "per-hour" numbers
are daily averages (reflected in the report heading).
--smtpd_warning_detail <cnt>
Limit detailed smtpd warnings reports to the top <cnt>.
0 to suppress entirely.
--syslog_name=name
Set syslog_name to look for for Postfix log entries.
By default, pflogsumm looks for entries in logfiles
with a syslog name of "postfix," the default.
If you've set a non-default "syslog_name" parameter
in your Postfix configuration, use this option to
tell pflogsumm what that is.
See the discussion about the use of this option under
"NOTES," below.
-u <cnt> top <cnt> to display in user reports. 0 == none.
See also: "-h" and "--*_detail" options for further
report-limiting options.
--verbose_msg_detail
For the message deferral, bounce and reject summaries:
display the full "reason", rather than a truncated one.
Note: this can result in quite long lines in the report.
--verp_mung do "VERP" generated address (?) munging. Convert
--verp_mung=2 sender addresses of the form
"[email protected]"
to
"[email protected]"
In other words: replace the numeric value with "ID".
By specifying the optional "=2" (second form), the
munging is more "aggressive", converting the address
to something like:
"[email protected]"
Actually: specifying anything less than 2 does the
"simple" munging and anything greater than 1 results
in the more "aggressive" hack being applied.
See "NOTES" regarding this option.
--version Print program name and version and bail out.
--zero_fill "Zero-fill" certain arrays so reports come out with
data in columns that that might otherwise be blank.
RETURN VALUE
Pflogsumm doesn't return anything of interest to the shell.
ERRORS
Error messages are emitted to stderr.
EXAMPLES
Produce a report of previous day's activities:
pflogsumm.pl -d yesterday /var/log/maillog
A report of prior week's activities (after logs rotated):
pflogsumm.pl /var/log/maillog.0
What's happened so far today:
pflogsumm.pl -d today /var/log/maillog
Crontab entry to generate a report of the previous day's activity
at 10 minutes after midnight.
10 0 * * * /usr/local/sbin/pflogsumm -d yesterday /var/log/maillog
2>&1 |/usr/bin/mailx -s "`uname -n` daily mail stats" postmaster
Crontab entry to generate a report for the prior week's activity.
(This example assumes one rotates ones mail logs weekly, some time
before 4:10 a.m. on Sunday.)
10 4 * * 0 /usr/local/sbin/pflogsumm /var/log/maillog.0
2>&1 |/usr/bin/mailx -s "`uname -n` weekly mail stats" postmaster
The two crontab examples, above, must actually be a single line
each. They're broken-up into two-or-more lines due to page
formatting issues.
NOTES
Pflogsumm makes no attempt to catch/parse non-Postfix log
entries. Unless it has "postfix/" in the log entry, it will be
ignored.
It's important that the logs are presented to pflogsumm in
chronological order so that message sizes are available when
needed.
For display purposes: integer values are munged into "kilo" and
"mega" notation as they exceed certain values. I chose the
admittedly arbitrary boundaries of 512k and 512m as the points at
which to do this--my thinking being 512x was the largest number
(of digits) that most folks can comfortably grok at-a-glance.
These are "computer" "k" and "m", not 1000 and 1,000,000. You
can easily change all of this with some constants near the
beginning of the program.
"Items-per-day" reports are not generated for single-day
reports. For multiple-day reports: "Items-per-hour" numbers are
daily averages (reflected in the report headings).
Message rejects, reject warnings, holds and discards are all
reported under the "rejects" column for the Per-Hour and Per-Day
traffic summaries.
Verp munging may not always result in correct address and
address-count reduction.
Verp munging is always in a state of experimentation. The use
of this option may result in inaccurate statistics with regards
to the "senders" count.
UUCP-style bang-path handling needs more work. Particularly if
Postfix is not being run with "swap_bangpath = yes" and/or *is* being
run with "append_dot_mydomain = yes", the detailed by-message report
may not be sorted correctly by-domain-by-user. (Also depends on
upstream MTA, I suspect.)
The "percent rejected" and "percent discarded" figures are only
approximations. They are calculated as follows (example is for
"percent rejected"):
percent rejected =
(rejected / (delivered + rejected + discarded)) * 100
There are some issues with the use of --syslog_name. The problem is
that, even with $syslog_name set, Postfix will sometimes still log
things with "postfix" as the syslog_name. This is noted in
/etc/postfix/sample-misc.cf:
# Beware: a non-default syslog_name setting takes effect only
# after process initialization. Some initialization errors will be
# logged with the default name, especially errors while parsing
# the command line and errors while accessing the Postfix main.cf
# configuration file.
As a consequence, pflogsumm must always look for "postfix," in logs,
as well as whatever is supplied for syslog_name.
Where this becomes an issue is where people are running two or more
instances of Postfix, logging to the same file. In such a case:
. Neither instance may use the default "postfix" syslog name
and...
. Log entries that fall victim to what's described in
sample-misc.cf will be reported under "postfix", so that if
you're running pflogsumm twice, once for each syslog_name, such
log entries will show up in each report.
The Pflogsumm Home Page is at:
http://jimsun.LinxNet.com/postfix_contrib.html
REQUIREMENTS
For certain options (e.g.: --smtpd_stats), Pflogsumm requires the
Date::Calc module, which can be obtained from CPAN at
http://www.perl.com.
Pflogsumm is currently written and tested under Perl 5.8.3.
As of version 19990413-02, pflogsumm worked with Perl 5.003, but
future compatibility is not guaranteed.
LICENSE
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You may have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307,
USA.
An on-line copy of the GNU General Public License can be found
http://www.fsf.org/copyleft/gpl.html.