[arch-general] Think twice before moving to systemd

[Felipe Contreras]

On Wed, Aug 15, 2012 at 7:35 AM, Tom Gundersen <teg at jklm.no> wrote:
> Felipe,
>
> On Aug 15, 2012 3:35 AM, "Felipe Contreras" <felipe.contreras at gmail.com>
> wrote:
>> I tried systemd a while ago in a brand new machine with Arch Linux and
>> the boot was *much slower*. After some exchanges with Lennart
>> Poettering and other people in Google+[1], it became clear I was on my
>> own. Eventually I found the culprit: Fedora uses CONFIG_HZ_1000, and
>> Arch Linux uses CONFIG_HZ_300. It became clear to me that systemd was
>> not ready for prime time, it wasn't thoroughly tested in a lot of
>> machines, and if you have problems Lennart Poettering will blame you
>> (PulseAudio sounds familiar?).
>
> Do you have a link to a proper bug report for this issue? I tried reading
> the Google+ thread but I couldn't stomach how rude you were in each of your
> messages (including the first one) so stopped reading.

No, I don't have such report.

>> systemd was the reason I stopped using Fedora in the first place; when
>> they moved to it my machine stopped booting reliably. My configuration
>> was non-standard (a single encrypted partition), so I guess they never
>> tested that. Similarly, I expect many Arch Linux users to bite these
>> corner-cases.
>
> Please note that we have waited much longer than Fedora did to make sure
> the corner cases have been taken care of. Is this problem still an issue,
> or is it just FUD? Link to (current) bug report?

I don't have that machine available at the moment, but I don't see how
such an issue could have been fixed given the lack of interest from
Lennart in that G+ post.

I'd say this issue most likely is not fixed, but it's only an example.
Just like one this there might be more.

>> Finally, it's much harder to debug. If you have a problem you will not
>> be able to open a script and figure out what is happening, and perhaps
>> modify it, and debug it. You would be greeted with an unmodified
>> binary, and the source code would be along these lines:
>>
>>
> http://cgit.freedesktop.org/systemd/systemd/tree/src/remount-fs/remount-fs.c
>
> As someone who has spent a lot of time debugging both, I much prefer
> systemd. I think you are being disingenuous here, surely you don't have a
> problem reading C?

And I prefer sh. Preferences don't count for much. I do read and write
C everyday for probably for more than 10 years now, yet I do have
trouble reading systemd's code, but that's not important, what is
important is that in order to test my modifications (to add debugging
for example), I would need to *recompile*.

>> I'm sure in due time systemd will be ready, and will have nice
>> advantages, but I doubt that's the case right now. Has anybody looked
>> into the CONFIG_HZ issue? I doubt that.
>
> This is the first I hear of it. I'd be interested to follow up if there is
> a proper bug report without unnecessary hostility.

Not to my knowledge.

>> I was expecting more from the Arch Linux community, something along
>> the lines of Google's analysis to pick to mercurial[2], but so far I
>> have only seen a couple of people saying +1 in the development mailing
>> list, with barely any explanation at all. Such an important move (one
>> that might make users' machines stop booting) should warrant at least
>> an analysis of some sort, with clear advantages. Would it not?
>
> We provided systemd optionally for a long time, as you know. Its pros and
> cons have been discussed at the various making lists at great length. A
> significant portion of our userbase has switched to it, and no serious
> issues seem to remain, based on the feedback we have been getting. Each dev
> will have had the possibility of trying it, and researching it. They will
> have done their own analysis on which the +1s are based. I see no value in
> providing an official public analysis. That's not how we work, and it would
> not help in the decision making at this point.

Well, I see absolutely no evidence of such an analysis, so consider me
a skeptic.

> That's not to say that an analysis would not be an interesting read, and
> I'm sure people like Allan our Jason will provide some excellent blog posts
> about this at some point.

One can only hope.

Cheers.

-- 
Felipe Contreras