Stan Hoeppner:
> Wietse Venema put forth on 9/27/2009 8:30 AM:
> > Stan Hoeppner:
> >> Wietse Venema put forth on 9/27/2009 7:39 AM:
> >>> If you have a question about POSTFIX logs, then it is a good idea
> >>> to send samples of POSTFIX logging.
> >> Point taken, sry Wietse. Upon grabbing these sample entries, I noticed
> >> the transactions pflogsumm is flagging have no qmgr entries. That
> >> explains why pflogsumm is complaining there is no size data. So, I
> >> guess the question is, why did qmgr intermittently/randomly stop
> >> stamping the log file, as of a day or two ago?
> >
> > Apparently, your system runs daemons in a chroot environment that
> > does not contain the required syslog socket.
> >
> > As shipped by me, Postfix does not run daemons in a chroot environment,
> > because the details vary with operating system and with system
> > version. If a maintainer or system administrator turns on the
> > Postfix chroot feature, then it is their responsibility to set
> > things up properly.
>
> Hi Wietse,
>
> If the chroot environment does not contain the required syslog socket,
> then why does qmgr place a stamp in the log for some message
> transactions and not others?
Oh, ye of little faith.
When you analyze the logs carefully, you will see that
1) qmgr logging starts after qmgr is restarted. Here, qmgr opens
the syslog socket BEFORE it enters the chroot environment.
2) qmgr logging stops after syslogd is restarted. Here, qmgr cannot
reopen the syslog socket, because there is no socket in the
chroot environment.
Now PLEASE go beat up the maintainers who ship Postfix in a broken
configuration. The more people compolain, the more likely that it
will get fixed,
Wietse
