irqbalance and at daemons by default?

[Daniel J Blueman]

On 26 March 2013 21:55, Chow Loong Jin <hyperair at ubuntu.com> wrote:
> On 26/03/2013 18:38, Daniel J Blueman wrote:
>> When setting up Ubuntu servers and desktops, two daemons I always
>> remove are the atd and irqbalance.
>>
>> irqbalance is perhaps good when you have a server with quad-port NIC
>> with a high rate of small packets and have found it to benefit over
>> the kernel's interrupt allocation (but wastes time and energy
>> otherwise); finally, the demographic who know and use the at daemon
>> must be >1% surely.
>>
>> What justification do we have for continuing forcing these on users by
>> default? (and can we win back some a slightly leaner, securer setup by
>> revisiting this logic?)
>
> I'm not sure you win much back:
>  - apt-cache show at | sed -n 's/^Size: //p' => 37376
>  - ps -C atd -o cputime,etime => 00:00:00 8-03:30:27
>  - ps -C atd -o rss => 68
>
> So all in all, 37.4k of disk space, 68k of memory, and approximately 0 seconds
> of CPU time out of 8d 3h.
>
> Are there many security vulnerabilities in atd?

The same can be said for irqbalance, except it does clock up cputime:

$ ps -C irqbalance -o cputime,etime => 00:40:55 82-02:24:34
...which is 30s/day on a single-user workstation

$ ps -C irqbalance -o rss => 392kb

But this is all missing a core tenet of Debian/Ubuntu: you select what
you want running and aren't imposed upon.

DL> at has reverse dependency lsb-core, that is if we care to support
DL> lsb-core set out of the box.

Out the box, lsb-base is installed; lsb-core isn't, so that doesn't
change anything (you'd still need to install lsb-core which would pull
in atd).

If we have no solid technical reasoning for imposing these daemons by
default, I'll propose we don't.

Thanks,
  Daniel
-- 
Daniel J Blueman