Victor Duchovni <[EMAIL PROTECTED]> wrote: > On Sat, Oct 11, 2008 at 11:03:37AM -0400, Sahil Tandon wrote: > > > Thank you and Viktor for your response. The sending MTA continues to > > retry on an hourly basis. I ran tcpdump as per the DEBUG_README and > > pasted one session capture (the output of tshark -n -r tcpdump_file) at > > the URL below. Prior to tshark, I tried viewing the tcpdump file with > > the following: tcpdump -nA -r tcpdump_file. However, the output of > > that commmand reveals the email as it was written, headers, etc. I wasn't > > sure that was necessary to debug the issue. But if it is, I'm happy > > to post that as well. > > > > I understand it is generally better to paste relevant excerpts in the > > body, but this particular capture is quite large! > > > > [TCP Previous segment lost] is followed by several duplicate ACKs, and > > eventually a 421 timeout error. > > > > http://pastebin.com/m7fb47518 > > Your server offers Window scaling (WS=3). Either the client's firewall or > yours is confused by window scaling. You may need to turn window scaling > off (it is a pity that many "new" TCP features are unusable in practice > due to firewall breakage).
That was it. After setting net.inet.tcp.rfc1323=0, mail is delivered to user's mailbox in less than 2 minutes after the initial connect. I'll take this to a more appropriate forum now, but want to understand whether disabling window scaling will adversely impact the system/postfix's performance. I tested sending much larger files from other servers and there is no problem even when scaling is enabled, so it appears the sending server is the one confused by scaling. -- Sahil Tandon <[EMAIL PROTECTED]>
