Reducing the load on Sysadmin

[Ben Cooksley]

On Mon, Nov 11, 2019 at 4:02 AM Luigi Toscano <luigi.toscano at tiscali.it> wrote:
>
> KDE ha scritto:
> > Hi,
> >
> > It seems there is always seem to be someone within KDE that wants something
> > new and shiny, I name gitlab, Discourse, a new identity system, etc.
> >
> > On the flip side, there is always someone that does not want to part with the
> > old stuff.
> >
> > Hence there is always more stuff to do, while we must also maintain all the
> > old stuff.
> >
> > Sometime you need a step back to create room to go two forward. We are just
> > asking to think with us if some services are really needed.
>
> This way of representing the reality is a bit of a stretch, to say the least.
>
> Most of the proposals haven't raised any concerns.
> No one is going to shed a tear on the demise of cgit.
> Moving to static-generator all the websites which does not need to be dynamic
> has been on the work for a lot of time, and it will continue.
> And so on.
>
> What's left? Some services which you may think have no users, but they really
> solve problems.
> API and EBN are probably not working in their current form, but something that
> covers they use case is needed anyway. There is some discussion, it's likely

Looks like we're missing part of the sentence here?

In any event, my email on that topic was mostly a state of fact that
the current way of operating it is unmaintainable and cannot continue.
It was a call of action as it were to the broader community to do
something about the problem in question.

Please note that nowhere in my mail did I mention shutting it down -
so i've no idea where this impression came from - because that is
certainly not what was proposed at all.

>
> Of course, when you propose many changes it is expected to see that not all of
> them may work as planned. That's why the discussion here is needed.
>
> About dealing with old services: it does not automatically mean "let's cancel
> them" but more "let's see if this is useful, and if it but we have workload
> issues, let's find a way to make it work".
>

As a rule, the approach i've taken does not take the age of the
service into account. We've been users of some software for many, many
years now (Drupal, Bugzilla, Mediawiki, etc) and have no plans to
discontinue them simply due to their age.

What I instead do is compare the benefit it provides to the community
as a collective whole versus the cost to provide it, and whether
something else we have within our systems can provide a similar
service.

To provide examples of how this logic works:

When assessing paste.kde.org, it was noted that we already had a
direct replacement for this, in the form of Snippets in Gitlab. As we
get this for free as part of running Gitlab, it means that the benefit
of running a separate Pastebin service now drops to effectively zero.

As the cost of maintaining paste.kde.org was now higher than the
benefit of providing it, it now makes sense to no longer provide this
service.

With regards to notes.kde.org, we were pointed in the direction of a
Nextcloud plugin which provided very similar functionality (with a
couple of minor regressions, but also improvements in some areas
including information security).

As it was rather costly to provide the notes.kde.org service
maintenance wise, and this had the effect of providing a very similar
service with a much reduced maintenance cost, it made sense to
discontinue the notes.kde.org service in favour of the Nextcloud
plugin.

> Talking about websvn, I'm pretty sure that there are various solutions and
> some of them have been proposed. To summarize the need: we do reference websvn
> from pointing out specific changes in emails or other channels. It can be done
> for git changes, it should still be possible for other changes.
>
> I has been said many times that recruiting new people for sysadmins is
> difficult for trust reasons, but I'm pretty sure that there are some people in
> the community who can be trusted by sysadmin right now. In fact there are
> already people managing some of the services, not all of them, so I don't see
> why it can't be done for other services.
> Or is the set of people who are potentially trusted by sysadmins really empty?
>

This is a two sided problem.
The first being that people aren't asking about looking after their
favourite service, the second being whether we can trust them.

To date we've not received much interest in regards to people joining.

Please note though that even if people do join, services must still
pass the benefit vs. cost test noted above (ie. even if available
human and technical resources were extended to infinity, the number of
things we would host would still remain limited)

>
> To be honest, the replies I've seen so far (which I'm nevertheless grateful
> for) looks more like point to point replies dismissing the proposals. What I'd
> like to see are more "yes, this can't be done this way BUT we may consider
> this.". Please try to help each other to find the solutions, keeping in mind
> that may not always fit your original plan.
>

Sorry i'm not sure if this is addressing Sysadmin or the people who
have been responding to our proposals.

With regards to WebSVN specifically, please note that what i've wanted
to find out here is exactly how it is being used, and what for
(because that is necessary to complete both the benefit vs. cost
analysis noted above, as well as to assess as whether an alternative
would be available to you)

> --
> Luigi

Cheers,
Ben