I am currently running a test to simulate a downed peer to see if the reg settings fix the problem, its a start to finding this problem that is killing me when a peer is down.
Rick Davidson
National Systems Manager
North American Title Group
-
----- Original Message ----- From: "Dmitri Elgin" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Thursday, September 23, 2004 10:08 AM
Subject: RE: [IMail Forum] D files with no Q files - Queuemanager crashes with unreachable peer (again)
I'm not sure that stopping queuemanager will prevent deliver to remote peer. It's a smtpd32.exe service task. So if you just stop and restart the queuemanager you just claim to redeliver all spool to the intendent recipients. In fact, when peer-to-peer connection established and D file being downloaded the Q file is already know by the server, since the connection filtering doing before downloading the mail content. But I don't know the reason why server create q file after downloading the appropriate d, so the situation is quite possible if the connection is broken.
Regards, Dmitri Elgin, http://imailzip.com
-----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Rick Davidson Sent: Tuesday, September 21, 2004 10:02 PM To: [EMAIL PROTECTED] Cc: [EMAIL PROTECTED] Subject: [IMail Forum] D files with no Q files - Queuemanager crashes with unreachable peer (again)
Is there a way to send orphaned D files that have no associated Q file other than manually creating a Q file?
Does anyone have a utility that will read through a directory of D files and create the Q files? Sure would be handy.
As I have posted many times in the past and submitted to Ipswitch, the problem with Imail Peering when a peer is unreachable (like during a hurricane) the master peer (the one behind the mail gateways) queue service will eventually fail with a buffer problem causing mail to not be delivered. A netstat reveals all the allocated ports in the TIME_WAIT state. I run seven peers across the country and occasionally an internet connection goes down making the peer unreachable. I was running a batch file to stop and
restart the queuemanager service once an hour (and recommended by Ipswitch support to work around the problem) I now realize that this is causing files to be orphaned due to the delivery process being stopped by my batch file, thus the D files without Q files.
Ipswitch claims the fault is with the MS TCP/IP Stack, hopefully the following info can generate some answers and some ideas from the brains on this list and at Ipswitch.
I found that the default application port allocation is from 1024 to 5000, when Imail attempts to reach a down peer and the queue is building up messages destined for the down peer these ports get used up sitting in the TIME_WAIT state. The busier the server the quicker the situation manifests itself.
The MS default TIME_WAIT is four minutes and there are 4076 ports available for Imail SMTP and Queue processes. I found the following MSKB articles showing reg hacks to modify the TIME_WAIT length and the default application port allocation.
I set the TIME_WAIT or Maximum Segment Life (MSL) (TcpTimedWaitDelay) to
half of the default time to two minutes.
I set the port allocation to 65534 (MaxUserPort)
http://support.microsoft.com/default.aspx?scid=kb;EN-US;196271 http://support.microsoft.com/default.aspx?scid=kb;EN-US;149532 http://support.microsoft.com/default.aspx?scid=kb;en-us;319504&Product=w in2000
I haven't had the buffer problem since but we haven't had a lengthy downtime to really see what happens and of course the last thing I want to do is hope for another hurricane to hit Florida just to test this :-) I will probly
have to setup a dummy peer and see what happens and that if it isn't fixed I catch it before the phones light up.
Any comments about these reg hacks would be welcomed especially if you know reasons why I should be doing this. Also it would seem to me that a mail
system should be able to grab more ports if it wants to be capable of extremely high loads without hitting a low port allocation wall and choking. It also makes me wonder if this is related to other queuemanager problems if the solution is to rake files out of the queue so it can catch up without stopping.
Rick Davidson National Systems Manager North American Title Group -
To Unsubscribe: http://www.ipswitch.com/support/mailing-lists.html List Archive: http://www.mail-archive.com/imail_forum%40list.ipswitch.com/ Knowledge Base/FAQ: http://www.ipswitch.com/support/IMail/
To Unsubscribe: http://www.ipswitch.com/support/mailing-lists.html List Archive: http://www.mail-archive.com/imail_forum%40list.ipswitch.com/ Knowledge Base/FAQ: http://www.ipswitch.com/support/IMail/
To Unsubscribe: http://www.ipswitch.com/support/mailing-lists.html List Archive: http://www.mail-archive.com/imail_forum%40list.ipswitch.com/ Knowledge Base/FAQ: http://www.ipswitch.com/support/IMail/
