Re: [exim] SMTP timeout after pipelined end of data (5334 by…

Page principale
Supprimer ce message
Répondre à ce message
Auteur: Marc MERLIN
Date:  
À: Evgeniy Berdnikov, Jeremy Harris
CC: exim-users
Sujet: Re: [exim] SMTP timeout after pipelined end of data (5334 bytes written), hang at DAT 7247 LAST. Local_scan vs chunking issue?
On Thu, Sep 10, 2020 at 10:47:45PM +0300, Evgeniy Berdnikov via Exim-users wrote:
> On Thu, Sep 10, 2020 at 12:24:51PM -0700, Marc MERLIN via Exim-users wrote:
> > But on the other side, I see:
> > 2020-09-10 12:05:33 1kGRsw-0006mt-VG SA: Debug: SAEximRunCond expand returned: '1'
> ...
> > 2020-09-10 12:05:34 cwd=/var/spool/exim4 3 args: /usr/sbin/exim4 -Mc 1kGRsw-0006mt-VG
> > 2020-09-10 12:05:34 1kGRsw-0006mt-VG => |/usr/bin/procmail -m /home/merlin/.procmail/procmailrc_marcme (marcatwww@???) <marcatwww@???> F=<SRS0=1t+S
> > Ht=CS=solar.cc=info@???> R=system_aliases_merlin T=address_pipe_root S=7550
> >
> > The Email was scanned in 1 second and accepted.
>
> Are you sure? There is no line indicating the end of message processing.
> I suppose it is somewhere below in the log, try to found it with
> "grep 1kGRsw-0006mt-VG".


Yes, I got 10+ copies of each Email from that domain.

On Thu, Sep 10, 2020 at 08:49:49PM +0100, Jeremy Harris via Exim-users wrote:
> On 10/09/2020 20:24, Marc MERLIN via Exim-users wrote:
> > I have an issue where some Emails go to the backup MX, then the backup
> > MX sends to the main MX, and the mail gets delivered to my mailbox, but
> > the transaction isn't finished in a way that the backup knows the mail
> > went through, so it sends it over and over again.
>
> What sort of proportion of messages?


Good question. It's quite rare, maybe less than 0.1% of messages?

> We'd really like to get debug from the primary MX for one such,
> or a packet capture (though we'd want the TLS session keys for
> the latter, which means debug on one end).


I can try, but indeed I can't reproduce this right now. After the last
queue flush, the message went through for real, and now everything is
good with the rest.

> By the way - you're advertising plaintext AUTH methods on
> plaintext connections. I *really* hope you don't accept
> any plaintext-conn AUTHs from outside your organisation
> (and, defence-in-depth, even within is poor practice).


Thanks for checking.
I am only using this for myself (single user) and only connect over
ssmtp.

Yes, I should probably reconfigure this later, but functionally it's the
same in my use case.

Thanks,
Marc
--
"A mouse is a device used to point at the xterm you want to type in" - A.S.R.

Home page: http://marc.merlins.org/                       | PGP 7F55D5F27AAF9D08