Re: qmail sighting

1999-08-17 Thread David Dyer-Bennet

Fred Lindberg <[EMAIL PROTECTED]> writes on 17 August 1999 at 13:31:15 -0500
 > On Tue, 17 Aug 1999 09:39:13 -0500 (CDT), David Dyer-Bennet wrote:
 > 
 > >I'm confused by the "completed messages" statistic; is the low value
 > >because it only lists messages for which all deliveries have been
 > >completed (a guess based on the name)?  Why does anybody *care* about
 > >such a bizarrely constrained statistic?  Also, how do you weight bytes
 > >by success?
 > 
 > >From matchup you have to collect fd5 output (pending messages) and feed
 > them in again to the next run of matchup (on the next log file). see
 > man page.
 > 
 > You're stats are screwed up because you didn't and there are a lot of
 > deliveries for which the initial log entry is not available to matchup.

I'm collecting the fd5 output, and I'm feeding it back in, really I
am.  Only for the last 3 or 4 days at this point, I'm just getting the
log rollover code to do matchup right.  There shouldn't be any
significant number of deliveries outstanding over multiple days,
though.  
-- 
David Dyer-Bennet ***NOTE ADDRESS CHANGES***  [EMAIL PROTECTED]
http://dd-b.lighthunters.net/ (photos) Minicon: http://www.mnstf.org/minicon
http://www.dd-b.net/dd-b (sf) http://ouroboros.demesne.com/ Ouroboros Bookworms
Join the 20th century before it's too late!



Re: qmail sighting

1999-08-17 Thread Fred Lindberg

On Tue, 17 Aug 1999 09:39:13 -0500 (CDT), David Dyer-Bennet wrote:

>I'm confused by the "completed messages" statistic; is the low value
>because it only lists messages for which all deliveries have been
>completed (a guess based on the name)?  Why does anybody *care* about
>such a bizarrely constrained statistic?  Also, how do you weight bytes
>by success?

>From matchup you have to collect fd5 output (pending messages) and feed
them in again to the next run of matchup (on the next log file). see
man page.

You're stats are screwed up because you didn't and there are a lot of
deliveries for which the initial log entry is not available to matchup.


-Sincerely, Fred

(Frederik Lindberg, Infectious Diseases, WashU, St. Louis, MO, USA)




Re: qmail sighting

1999-08-17 Thread David Dyer-Bennet

Dave Sill <[EMAIL PROTECTED]> writes on 17 August 1999 at 08:16:42 -0400
 > Bruce Schneier's CRYPTO-GRAM, a monthly newsletter sent to over 20,000 
 > subscribers, uses qmail and ezmlm. This a high-profile list and a
 > juicy hacker target.

Yep, it does.  With regard to which, I'm having trouble understanding
the qmailanalog statistics I get out of this.

Crypto-gram was sent yesterday in the early evening.  As of now, I
have only 20 messages in the queue.  There was a tiny "pending" file
generated by matchup last night, and only one of the messages was
crypto-gram related (so most of the deliveries had to happen before
the log cutoff last night).  Here are the statistics on yesterday and
today so far (zoverall):

Basic statistics

qtime is the time spent by a message in the queue.

ddelay is the latency for a successful delivery to one recipient---the
end of successful delivery, minus the time when the message was queued.

xdelay is the latency for a delivery attempt---the time when the attempt
finished, minus the time when it started. The average concurrency is the
total xdelay for all deliveries divided by the time span; this is a good
measure of how busy the mailer is.

Completed messages: 1620
Recipients for completed messages: 4971
Total delivery attempts for completed messages: 5126
Average delivery attempts per completed message: 3.1642
Bytes in completed messages: 15183142
Bytes weighted by success: 31852134
Average message qtime (s): 197.403

Total delivery attempts: 25933
  success: 24576
  failure: 289
  deferral: 1068
Total ddelay (s): 66248917.435193
Average ddelay per success (s): 2695.675351
Total xdelay (s): 391021.65
Average xdelay per delivery attempt (s): 15.078137
Time span (days): 8.0081
Average concurrency: 0.565141

I'm confused by the "completed messages" statistic; is the low value
because it only lists messages for which all deliveries have been
completed (a guess based on the name)?  Why does anybody *care* about
such a bizarrely constrained statistic?  Also, how do you weight bytes
by success?

A quick check with qmail-qread shows that indeed there are a very few
pending deliveries on crypto-gram.  When those finally clear out, one
way or the other, will I suddenly get an *immense* lump in my
"recipients for completed messages" that day?  As I say, I think it's
a silly number to compute.

(concurrencyremote on this system is 50; it's a Cyrix 166 running
Linux with 96 meg of ram, IDE disks, no RAID.  I'm amazed how well it
digests big lumps like the crypto-gram mailing.  A normal day here is
a couple thousand deliveries.)
-- 
David Dyer-Bennet ***NOTE ADDRESS CHANGES***  [EMAIL PROTECTED]
http://dd-b.lighthunters.net/ (photos) Minicon: http://www.mnstf.org/minicon
http://www.dd-b.net/dd-b (sf) http://ouroboros.demesne.com/ Ouroboros Bookworms
Join the 20th century before it's too late!



qmail sighting

1999-08-17 Thread Dave Sill

Bruce Schneier's CRYPTO-GRAM, a monthly newsletter sent to over 20,000 
subscribers, uses qmail and ezmlm. This a high-profile list and a
juicy hacker target.

>To subscribe, visit http://www.counterpane.com/crypto-gram.html or send a
>blank message to [EMAIL PROTECTED]

-Dave