From: Wietse Venema (wietseporcupine.org)
Date: Tue Mar 06 2007 - 08:04:26 CST

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Ralph Seichter:
[ Charset ISO-8859-1 unsupported, converting... ]
> Lately, I keep seeing log message pairs like the following:
>
> Mar 6 04:43:31 host postfix/postdrop[15763]: warning: uid=0: Illegal seek
> Mar 6 05:43:31 host postfix/sendmail[15760]: fatal: root(0): queue file write error
>
> The apparent one hour gap does not really exist, these messages appear
> on consecutive lines of the mail log.

As far as I know, postdrop copies stdin to queue file without any
need for seeking.

It is possible that a buggy file system mis-reports file size errors
as seek errors.

What file system is this?

        Wietse

> At 05:43 local time, a cron job generates a large binary file and uses
> nail to send the file as an attachment, by dint of Postfix's sendmail.
> This file's size has grown over time, so I suspected a connection. When
> I increase message_size_limit sufficiently, the log entries no longer
> occur.
>
> Can Postfix be configured to allow the size of sendmail-injected mail to
> exceed the general message_size_limit? Can the fatal queue file write
> error perhaps be avoided in future Postfix releases?
>
> Please let me know if you require "postconf -n" output.
>
> -R
>
>

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]