reopen 360873 tags 360873 - sid kthxbye On Tue, Sep 12, 2006 at 04:03:53AM -0700, Debian Bug Tracking System wrote: > Exim4 already sends multiple messages in the same connection if > possible. See: > http://www.exim.org/exim-html-4.62/doc/html/spec_html/ch30.html#id2645371
I seem not to be having further problems with spohr (or now, rietz), so I consider 367665 to be done. However, murphy still causes a huge number of connections. This is probably because murphy runs postfix: sh-3.1$ nc murphy.debian.org. 25 220 murphy.debian.org ESMTP Postfix (Debian/GNU) This makes what exim does rather irrelevant. :-) Please also note that I am subscribed to exactly one open (non-announce) list on murphy: debian-l10n-english. Think about when I was subscribed to -devel, -project, and -legal. ;-) > What is most likely happening is that you've got ClientRate set too lo > in your access file, which is running into the Rate limiter well > before it hits the connection limiter. sh-3.1# host murphy.debian.org. murphy.debian.org A 70.103.162.31 sh-3.1# grep -C1 70.103.162.31 /etc/mail/access # I have moved GreetPause 1000->30000, and ClientConn 30->1. bmc 2006-09-22. GreetPause:70.103.162.31 30000 ClientRate:70.103.162.31 30 ClientConn:70.103.162.31 1 # The window size is 60 seconds, so murphy is permitted to make up to 30 connections, one at a time, in that period. Due to a number of factors, including bandwidth limitations, no remote mail is ever delivered in less than three seconds; therefore, murphy could not *possibly* hit the rate limiting if it reuses connections. Note that as of 2006-09-22, I imposed additional restrictions on murphy because it couldn't comply with the original, much more lenient restrictions. > Disable the rate and/or connection limiter for spohr and murphy and > you'll stop seeing the problem. Uh, murphy ran into the global limits originally, which is why I filed this bug. It then ran afoul of my more generous limits. I'm not going to disable limiting connections for a mail server which abuses my generosity and cannot comply with what has been common practice for at least five years. There is *absolutely* *no* *reason* why murphy cannot reuse the connection; when it does, the rate limits will become moot. Just to prevent this already long email from getting longer, I have placed the relevant portions of the logs (in UTC and including portions from rietz, for comparison) at <http://crustytoothpaste.ath.cx/~bmc/files/free/dump/murphy-connrate>. I'm not trying to be nasty, so please don't take it that way; I'm just a bit frustrated. I'm trying to make my logcheck results a little smaller, so I have more time to work on those FTBFS-with-gcc-4.[23] bugs. :-) -- brian m. carlson / brian with sandals: Houston, Texas, US +1 713 440 7475 | http://crustytoothpaste.ath.cx/~bmc | My opinion only Screw you, AACS LA: 09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0 OpenPGP: RSA v3 2048b 560553E7: FE82 7C9F EB21 5436 2F96 25BA 927B 0A51 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
