2.7 Pitfalls
Before replacing your current
sendmail with a new version, be sure that the
queue is empty. The new version might not be able to properly process
old (or different) style queued files. After running the new sendmail for
the first time, look in the queue directory for filenames that start
with an uppercase Q, which can indicate a problem.
See Section 11.5 for a description of why these files
appear and what to do about them.
If you change the location of the queue to a different disk, be sure
that disk is mounted (in /etc/rc) before the
sendmail daemon is started. If
sendmail starts first, there is a risk that
messages will be queued in the mount point before the disk is
mounted. This will result in mysteriously vanishing mail.
Always save the old sendmail and configuration
file. The new version might fail when you first try to run it. If the
failure is difficult to diagnose, you might need to run the old
version while you fix the new version. But beware that the old
version will probably not be able to read the queue files created by
the new version.
Some operating systems allow disks to be mounted such that
set-user-id permissions are disallowed. If you
relocate sendmail, avoid locating it on such a
disk.
Don't be mistaken in the belief that
nis will correctly give you MX (Mail eXchanger)
for hosts. If, after compiling and installing
sendmail, you find that you cannot send mail to
hosts using MX records, you should recompile with NAMED_BIND defined
(NAMED_BIND). Also note that a misconfigured
service-switch file can also prevent proper MX lookups (ServiceSwitchFile).
|