Luke Crawford wrote:
heh. You are talking to a person that has maybe 1/5th the value of his
vehicle in safety gear. I do put quite a bit of effort into protecting
myself from myself... From my perspective, nobody else is quite as
dangerous.
Clearly safety is important; thus the wide adoption of version control
tools, unit testing, etc.
However, those are ways of making the core, the language, more safe.
I'm not complaining about safety -- I'm complaining about people
choosing crappy, ancient tools and techniques because they're "safe" (as
if any newbie with a root password is "safe").
certainly, our respective credentials speak for themselves- I'm not the
expert here, but for what it's worth, my experience has been that the
tool that presents a low barrier to entry wins; I believe Unix and C
win because they are simple, not because they are powerful. Give me any
reasonably intelligent person for a month and I'll give you back a
mediocre unix admin.
I agree that simplicity is probably one of the biggest factors, but
artificially weak tools tend to get rejected. At least, I hope,
especially since I'm developing all of my tools based on this assumption.
more recently Java (and Object oriented languages in general) won,
again because of the low barrier to entry (and this time at the expense
of simplicity.) - OO enables you to throw a bunch of mediocre
programmers at a programming job, and to get something bloated and
horribly slow out the other end; but something that mostly works. With
C, the quality of your programmers needs to be another order of
magnitude higher.
People seem to be quite capable of making horrendous programs in any
language you throw at them.
Well, yes. See, I am the boss at the place where I'm allowed to play
with system management tools. Unfortunately, this being the real world,
and me supporting it off of my consulting activities, I can't afford a
SysAdmin who is both good at figuring things out, independent, and
careful. I actually have one of each at the moment, and I'm still
working on getting them to work as a coherent unit. (And no, I'm not
going to fire him for botching a mdadm... I'm not paying that much more
than retail here; I have to expect that he will be doing some learning.)
I like to think that with tools like Puppet you could get more sysadmin
for your money as long as your people are capable of learning. That is,
if you started your sysadmins using these higher-level tools, rather
than starting with scripting, you might get a greater reward.
My setup right now is "powerful and dangerous" - everything is on a
SAN, and I'm using software raid. It's fairly easy to destroy a whole
lot of customer data. (in fact, that's what happened last weekend;
after a reboot (due to a homicidal oom_killer) the md devices didn't
come up correctly, and someone forced them up, but swapped the devices
around; we now have two functional mirrors with the same data; all
data on mirror 2 is now gone.)
Ouch.
Frankly, I've learned far more from doing stupid things with tools
than I ever did from doing smart things. I'll give you another quote:
Good judgment comes from experience, and experience comes from bad
judgment. --Barry LePatner
That is a good point when it comes to choosing a training tool.
Heh, all of life is a training tool. :)
systematic problems are personal problems if you hired someone else to
run your networks.
I wish business owners were better at recognizing competence as well. I
can cite many examples where people that I knew were obviously better
than I am got passed up, when I got hired. Some of these times I have
even told the hiring person "hey that guy over there- he's obviously
better than I am" (that is usually after I have the job; still, I have
access to good cheap people- I give them away to my employers, and my
employers prefer to find their own failure.
The reality most of us are dealing with, though, is one of mediocrity.
I definitely agree with that, but it just seems insane to build for the
trailing edge instead of the leading edge.
Me, I attempt to deal with this using "extreme redundancy" - but that
is expensive. (but can be cheaper than good sysadmins)
I love the idea of more agile, on-demand system administration in place
of the huge, redundant systems of the past.
the truth is that humans are failable; some more than others. anything
you can do to mitigate this is good. Even for me, I know I make
mistakes. Extreme awareness of my own failability helps a whole lot;
but most employers select based on confidence, so you certanly can't
count on that.
I'd say most employers select based on fear, but I'm completely
uninterested in building tools for the fearful, there are already plenty
of people targeting that demographic.
almost. I tell people to learn C, then a lisp/scheme, then whatever
the current buzzword language is; Once you know c and lisp, everything
else is a weekend of hacking to figure out. Putting all your eggs in
one buzzword is a bad idea, I tell them.
I don't really care about the language sequence; my point is that
non-portable scripting is the assembly of the sysadmin world, and
telling people to start with it is like telling programmers they should
work in assembly for a couple of years. The experience just doesn't
translate unless you're writing device drivers or compilers.
and again; I'm not trying to say that your tool is worthless or that
I'm better than you; I am clearly, as the kids would say, the n00b
here. I'm just trying to give you a view from what is perhaps a
different part of the SysAdmin world than you have experienced.
I understand that. You're working with the unfortunate reality of tools
and practice today, though, and I think it's a mistake to design tools
to that state, rather than to where we could be. At the least, I think
the programming world sets a very good model to use in designing tool
ecosystems.
--
The Number 1 Sign You Have Nothing to Do at Work...
The 4th Division of Paperclips has overrun the Pushpin Infantry
and General White-Out has called for a new skirmish.
---------------------------------------------------------------------
Luke Kanies | http://reductivelabs.com | http://madstop.com
_______________________________________________
lssconf-discuss mailing list
[email protected]
http://lists.inf.ed.ac.uk/mailman/listinfo/lssconf-discuss
