Default-installed MTA (was Re: MTA virtual provides craziness)

Nico Kadel-Garcia nkadel at gmail.com
Tue May 21 00:20:56 UTC 2013


On Mon, May 20, 2013 at 4:33 PM, Reindl Harald <h.reindl at thelounge.net> wrote:
>
>
> Am 20.05.2013 22:27, schrieb Peter Robinson:
>> You still have to configure it so doing a "yum install
>> your-mta-of-choice" isn't hard and in fact in all environments I know
>> of that have auto monitoring and alerting of drive failures configure
>> it automatically as part of a kickstart or puppet deployment and use
>> snmp rather than email to deal with that and have active alert
>> systems
>
> you refer to enterprise environments
> i live in both, real enterprise and SOHO
>
>>> if a disk dies it is nice to have it in syslog but
>>> it is useless if you see it days later while a mail
>>> from crond is more or less real time
>>
>> You still have to configure all of that and whether a MTA is installed
>> automatically or not doesn't really make it work out of the box.
>
> you have *ntohing* to configure
> you only need to edit *one line* in /etc/aliases
> everybody who knows unix-like systems knows this

I'm afraid not. You have to make some other choices.

If your envornment uses outbound port 25 blocking, and particularly
enforces a SMARTHOST, either you allow /etc/aliases (with sendmail)
but give up auto-forwarding of "local" email to the SMARTHOST, or you
use Postfix and /etc/aliases and $HOME/.forward are ignored. The only
MTA I've seen handlie both SMARTHOST and /etc/aliases correctly was
"exim",. which is no longer included by default and for which the
necessary configuration is not widely published nor in the default
documentation.

There's also the "sendmail" problem of having the system's output of
"hostname --fqdn" be listed in /etc/hosts as the first entry, or
reliably available in DNS at boot time, in order to start the sendmail
daemon without a galling 5 minute wait that is just nasty in a
production environment.


More information about the devel mailing list