Another aspect of "the change" is the proliferation of "cloud
services". If you are on the consumer end, the tools are of one type
while the providers use a totally different type.
of) nodes of a donated platform with limited storage and bandwidth. The
nodes are not joined by by interconnect processes like MPI. Jobs must
after each run to get as much done as possible. I don't have root
streams into RAM from networked sources.
Yep. Different kind of problem yet still very similar - automate a
complicated task at scale with essentially minimal tools at hand.
cool stuff and I am NOT complaining at all. Unlimited resources
decreases the challenge and thus the fun.
Post by Jerald SheetsBash scripts over ssh. Nothing says "my network and server farm"
like customs solutions for specific problems. Add a custom pack of
cool tools on a server build out.
Post by DJ-PfulioPuppet was the early winner. Now it feels like picking MSFT solutions.
Nobody gets fired for going with the most popular
solution. You'll
always be employed, but might hate it.
TL;DR: Everythingâs in a bubble right now, all the tools are about to
change, and the methods typically assigned to the sysadmin will need
to look fundamentally different while the sensibilities need to
remain the same and a new culture needs to be apprehended and
embraced.
And these are the conversations weâre having in the DevSecOps world
and in the Sysadminâs groups (USENIX, LOPSA, et. al.)
We have a ton of âcool toolsâ weâve either assimilated or built
ourselves over our collective careers, and we have our own
denomination in the systems world. Thing is, the job hunts any more
are requiring skillsets that have Config Management right square in
the middle of them, and Chef/Puppet are the clear winners there.
A lot of recruiters are lumping Ansible in with those two, but there
is a difference between a DSC paradigm and a âpush what I need out
thereâŠeverywhereâ paradigm. Iâve always separated Ansible into its
own world of which it is the master of that domain, and that is one
of orchestration. Puppet and Chef are nowhere near as good at
orchestration as Ansible, and is why we have it paired with Puppet
for the workload we have. Neither is up to the task in and of
itself, but needs the capabilities of the other to be a powerful,
wholistic toolset.
Anywho⊠I talk to recruiters all the time who canât define âDevOpsâ.
The toolset that could constitute a DevOps is so vast and arrayed,
finding people whoâve taken the time to start digging into all of it
is very difficult. Thatâs why the more capable of the DevOps-ians
are making 160-200k in Atlanta. The problem is that âDevOps
Engineerâ in Atlanta also makes 80k. Just search the boards⊠itâs
confusing and can be maddening.
Now what do we do? How do you define these things such that it makes
the hiring process easier? This is becoming a problem because DevOps
(and by implication, Configuration Management) has such a WIDE
description and salary range that the exact same term at two
different companies can mean the difference between feast or famine,
entry/mid level or senior architect, and unfortunately means two
entirely different toolsets and two entirely different skillsets.
Iâve always been one to try and stay one step ahead of the âcoming
waveâ and be tooled to be marketable for âwhatâs nextâ. Ask Joey
Kelly right here on this board⊠heâs been the same way since the
early 90âs when I first met him. I remember when no one else was
talking about it, he was a firm proponent that biblical translation
work should happen on CD technology (this was before CDs were
released as data storage mechanisms, btw) when translators were in
the field to avoid loss of translation work and resiliency against
the elements. Wycliffe translators (largest translation group in the
world) started using that as the standard toolset 5 years later (an
eternity in computer-speak), and have now moved to removable media of
the USB drive ilk. In Baton Rouge, I was Linux before Linux was a
thing. I installed the first Slackware at LSU from floppies mere
months after Linus had announced the kernel and slack had produced a
distribution. I like to stay ahead. :)
The issue isnât Puppet or Chef, Ansible or SALT, and it isnât whether
youâre Linux/UNIX or Win, small environment or large.
The issue is CULTURE. 20 years ago, the very idea of a tool that
could manage the configuration of âall the thingsâ was anathema to a
Systems person. We had our scripts, our Perl, our host files, and
our provisioning nexxi, and we would be damned if we were going to
change. Fast forward, and we still have our tools and our methods,
and its âgood enough for what Iâm doing and Iâll be damned if Iâm
going to changeâ is still our story, but I fear not for whatâs
adequate for today, but whatâs necessary for tomorrow.
I submit to you that the current trend is that you MUST have a config
management background to be employable any more, and the better you
grasp the dev methods out today, apply them to infrastructure ,and
become portable in that knowledge, the further up the salary chain
you will find yourself. âLinux Adminâ as a discipline is seeing its
salaries normalize and in some cases go down. Way down. If you want
sooth elder years, you might want to rethink strategy, as the old
methods and tools might find you unemployed in 5-10 years.
Problem is, as you all know, this is a bubble (the config management
one, anyhow) and itâs coming to an end! IAC (Infrastructure as Code)
is sticking around, but new tools and methods are emerging that are
better at âeven biggerâ where tools like Puppet and Chef (and to some
degree Ansible, but not as much because itâs not like those other
kids) are starting to fail. And that is massive scale. Tools like
Terraform, Consul, Vault, and Nomad are designed for massive scale
and were essentially built in the cloud. Both Chef and Puppet are
being bolted onto the cloud and cloud methods.
Before you cry âfoulâ, both Adam Jacobs and Luke Kanies (creator of
Chef and Puppet respectively) feel this way, and theyâve said as much
to my face, that they wish their tools were more cloud friendly, and
the biggest challenge they will need to solve is to become more
âcloud friendlyâ as the months go on.
So then, why this missive in the middle of such a conversation?
Administration is changing, we have to as well.
Scale is going to be king and while you may not need scale now, you will.
Our discipline is morphing, and the âcool toolâ of now will debut and
wane, but DevOps culture and IAC sensibilities will never die.
Iâm not worried about what you need today, but what you WILL need
tomorrow next time you heat up your resume at Dice, Indeed, or
whatever the new tool will be in the future.
The conversation is bigger than the tools.
_______________________________________________
Ale mailing list
http://mail.ale.org/mailman/listinfo/ale
See JOBS, ANNOUNCE and SCHOOLS lists at
http://mail.ale.org/mailman/listinfo