[fetchmail]Fetchmail daemon mode
Matthias Andree
matthias.andree@gmx.de
Wed, 20 Apr 2005 03:56:10 +0200
"Bill Greganti" <t20030101@greganti.com> writes:
>> -----Original Message-----
>> From: fetchmail-friends-admin@lists.ccil.org
>> [mailto:fetchmail-friends-admin@lists.ccil.org] On Behalf Of
>> Matthias Andree
>> Sent: Sunday, April 17, 2005 12:36 PM
>> To: fetchmail-friends@lists.ccil.org
>> Subject: Re: [fetchmail]Fetchmail daemon mode
>>
>> > Actually, the daemon runs as the user fetchmail by default, not as
>> > root.
>>
>> Is this something you're doing explicitly? Just running fetchmail as
>> root user will not drop privileges.
>
> That's seems to be the default for the Debian package. I haven't
> changed anything.
Then the Debian guys must've poked the code. I don't have a "fetchmail"
user on my non-Debian system.
Please don't assume $DISTRIBUTOR's behavior is the same as default.
> The would require me to make a cron job for each user that uses
> fetchmail, unless I'm missing some clever way to automate this.
I wonder if some of the service dispatchers can handle such tasks, there
are cron variants, runit, uoschedule, but I don't know about their
capabilities.
>> You don't want to do that on very old operating systems or if you're
>> short on memory.
>
> The machine is an Athlon XP 2200 with 1GB of memory. I guess
> I'm just a little too used to working with Windows machines where
> 10 processes running crashes the machine. :)
If you get those 10 processes up before work is over or you need to go
to bed. :)
>> A server that causes timeouts delays mail for all other
>> users, with many
>> servers to poll up to the point where a single run of that single run
>> control file takes longer to process than your users are willing to
>> accept.
>
> Doh! With thousands of users that could pose a real problem in the
> future. It becomes too likely that one server will gum up the whole
> process at some point.
Indeed, which is why separate configuration files, one per user, or
grouped by upstream server at most, are my recommendation.
--
Matthias Andree