From owner-freebsd-hackers Tue Dec 19 01:17:01 1995 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id BAA02551 for hackers-outgoing; Tue, 19 Dec 1995 01:17:01 -0800 (PST) Received: from mail.barrnet.net (mail.barrnet.net [131.119.246.7]) by freefall.freebsd.org (8.7.3/8.7.3) with ESMTP id BAA02545 for ; Tue, 19 Dec 1995 01:16:57 -0800 (PST) Received: from haywire.DIALix.COM (haywire.DIALix.COM [192.203.228.65]) by mail.barrnet.net (8.7.1/MAIL-RELAY-LEN) with ESMTP id LAA17247 for ; Tue, 12 Dec 1995 11:26:01 -0800 (PST) Received: (from news@localhost) by haywire.DIALix.COM (sendmail) id DAA08741 for freebsd-hackers@freebsd.org; Wed, 13 Dec 1995 03:23:13 +0800 (WST) Received: from GATEWAY by haywire.DIALix.COM with netnews for freebsd-hackers@freebsd.org (problems to: usenet@haywire.dialix.com) To: freebsd-hackers@freebsd.org Date: 13 Dec 1995 03:12:41 +0800 From: peter@haywire.dialix.com (Peter Wemm) Message-ID: <4akk79$89r$1@haywire.DIALix.COM> Organization: DIALix Services, Perth, Australia. References: , <199512111122.MAA08607@caramba.cs.tu-berlin.de> Subject: Re: mail storm Sender: owner-hackers@freebsd.org Precedence: bulk wosch@cs.tu-berlin.de (Wolfram Schneider) writes: >Jonathan M. Bresler writes: >>On Sun, 10 Dec 1995, Wolfram Schneider wrote: >> >>> Christoph P. Kukulies writes: >>> >Overnight I had received 650 mails or so (instead of 120 - my >>> >normal daily load) and many of these are deja-vu mails. >>> >>> Me too. I detect funny ``Received: '' lines, 550 mails via ra.dkuug.dk >>> and 100 mails via th-darmstadt.de. >> >> send me one, including headers please, if you have any that you >>have not deleted >I received 2760 Mails since november 29. 83 Mails are dups, mostly >cross postings (cc: hackers, current). 10 Mails seems really dups, but >10 Mails is not a problem for me ... >Wolfram I have an explaination for the volume and a theory about the duplicates.. The basic problem was that due to bad connectivity from freefall to europe and other places, there was a MAJOR pileup of queued mail on freefall. There were nearly 1000 emails to *.de sites that had been sitting there for 5 days, and were just about to be bounced to the postmaster on freefall (5 day timeout). I split the mail queue from one "deferred" queue on freefall to 5 queues, one for each day of backlog. After making arrangements with a FreeBSD core member in Denmark (Poul-Henning Kamp), *all* of the backlogged mail was sent to a machine under his control for exploding and delivery to the *.de and other north-eastern european sites. This would account for the massive flood of email. You could have receieved as many as 900 to 1000 emails over a few hours. Also, there is an race condition in the SMTP protocol that is tickled on bad internet links. Picture this: The originator (freefall) writes out the message and the trailing "." to end the body, and waits for the response for a few minutes. If it doesn't get a response, it times out and requeues the message... *however*, the network may be slow, and the final 10 or 20K of data including the "." may take a few minutes to arrive, and the numberic response code may be delayed due to the pathological TCP retransmit backoff.. But in reality, the remote machine receieved the mail via SMTP and responded, but freefall had given up waiting. At this point, there is now a duplicate mail in the system..... Considering the sheer volume of mail sent, and the current extreme packet losses across international links, I suspect it is most likely a manifestation of the SMTP race condition. We can only fix that by lengthening the SMTP transaction timeouts, which will cause freefall's mail queue to suffer even more in the face of the numerous genuine broken mailers out there, that genuinely timeout. If the problem persists, we may have to try something... -Peter