Turns out this is a general sendmail issue, not SUN-specific, but I
thought the summary might come in useful anyway.
The original posting was:
> We are having a problem with sendmail on our system here. We have a
> SPARCstation 5, Solaris 2.3, kernel patch 101318-54, sendmail patch
> 101739-04. This machine is connected to our ISP via PPP and a 14.4k
> modem.
>
> The ISP is trying to send us a huge (about 5MB) mail message that is
> in it's mail queue. Once about 400K of this message have been
> received into /var/spool/mqueue, sendmail exits and logs the message:
>
> Nov 10 14:20:35 neds-point sendmail[1825]: OAA01825: SYSERR(root):
> collect: I/O error on connection from sundog.tiac.net,
> from=<jeelliot@hookup.net>: Interrupted system call
>
> Unfortunately no response is sent back to the sender indicating that a
> problem occurred, so after a few minutes they try and send us the
> message again, with the same results. This is eating up our already
> limited bandwidth. Has anyone experienced this problem? Is there
> some kind of limit to the size of messages that can be received? We
> have plenty of space on disk for the mail queue.
>
> I couldn't find any documentation on this error in the sendmail book,
> which has otherwise been very comprehensive.
I got two responses which led me to the solution.
------------------------------------------------------------
Barry Margolin <barmar@bbnplanet.com> wrote:
The ISP is probably timing out waiting for the response saying that you
received the message. 400K takes about 5 minutes to transfer at 14.4 Kbps;
that's not an unlikely timeout. Ask your ISP to increase their timeout (I
think it's a sendmail config file option); for 5MB you'll need at least an
hour.
------------------------------------------------------------
Tony Lorimer <tlorimer@au.mdis.com> wrote:
I had a very similar problem here runninf sendmail-8.6.12. Have a look
at the timeouts in your sendmail.cf. You need to increase the
read timeout. You must try and guess how long it will take 5MB to
get down over your 14.4K link and also take into account all your
other traffice on the link as well.
------------------------------------------------------------
Bingo! I just needed to change the following line in my
/etc/mail/sendmail.cf:
WAS -
# read timeout for SMTP protocols
Or15m
NOW -
# read timeout for SMTP protocols
Or2h
With a two hour timeout I had no trouble receiving the files. I
didn't need to contact my ISP; their timeout was already large enough.
Thanks to Barry and Tony and apologies to any whose responses arrive
after I send this summary.
Ian
-- Ian Fox Senior Software Engineer Applied Language Technologies 215 First Street Cambridge MA 02142 Phone (617) 225-0012 Fax (617) 225-0322
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:35 CDT