[Exim] Desperation: Connection Reset by Peer

トップ ページ
このメッセージを削除
このメッセージに返信
著者: Martin W
日付:  
To: exim-users
題目: [Exim] Desperation: Connection Reset by Peer
Hello,

I'm nearly at breaking point with this so any advice on where to go now
would be welcome.

In a nutshell - sending messages with attachments from our mail hub to
certain sites will consistently fail with a "connection reset by peer".
One of these sites is BT Internet, and another Yahoo, so I'm assuming it's
a problem at our end. Sending smaller messages (ie no attachment) works
fine. Sending to the majority of other sites is also fine. The attachments
only have to be around 700K to cause the failure.

We're running exim 4.34 on a fast (dual Xeon/gig-E etc etc) Linux
(RH9 - 2.4.20-31.9smp) box.

I've tried all the suggestions here:
http://www.exim.org/exim-html-4.00/doc/html/FAQ_0.html#TOC17
and no go.

A tcpdump shows that the delivery all goes swimmingly until half way
through sending the attachment when, suddenly, the remote peer sends a TCP
reset out of the blue.

Another bizarre observation occurred when we were trying to monitor
traffic at the border of our network. We placed a 100M *hub* between the
router and the external link in order to 'tap' the connection. I know this
is unwise but, as I mentioned, we were desperate. All the time the hub was
in place, the problem went away. We could send mail happily to all of the
sites we had problems with previously. Heisenberg eat your heart out.
Removing the hub, the problem came back.

No doubt the hub was a bottleneck and was causing retransmits etc
effectively throttling the rate we were sending.

One other possibility is that I think all of the problematic remote sites
are running FreeBSD. I may be wrong about this tho'.

Any thoughts or advice would be most welcome. Help!

Regards,
Martin