Fri, Dec. 7th, 2012, 04:31 pm
the problem with devops
Gene Kim is a guy who's accomplished a lot more in his career than i have. I kinda-sorta met him in passing at an LSPE Meetup and he seemed like a nice guy. But his recent writeup, "Why We Need DevOps Now", which title i agree with, is just not good at all (i used the word "terrible" on Twitter, which someone called me on, so maybe i'll try to dial back the hyperbole).
As my friend John Willis told me after I dismissed DevOps as just another marketing fad, “DevOps is the best chance at relevance that IT Operations has had in thirty years.” I immediately realized that he was right.
Is that really all it took?"John, it's a fad."
"Gene, it's our last, best hope."
"OMG UR RITE!!"
DevOps is real (this article
says everything that needs to be said about it) and also a fad
, in the same way that Agile Software Development is real
and a fad:
BOSS: "We're going to try something called agile programming. That means no more planning and no more documentation. Just start writing code and complaining." WALLY: "I'm glad it has a name." BOSS: "That was your training."
A good business avails itself of forward-looking approaches in order to contend with the firehose of change that is our industry's lifeblood. But we cannot mistake these approaches for anything other than a tool.
Act I begins with IT Operations, where we’re supporting a large, complex revenue generating application. The problem is that everyone knows that the application and supporting infrastructure is... fragile.
I smell a setup. The scenario described is clearly the result of bad management, who failed to see this sort of undesirable performance down the road and act upon it to preclude the fragility, and it has been that way every time i've encountered it in my career. It is not, as implied, a result of traditional IT operations.
In Act 2, our life gets worse when the business starts making even bigger commitments to Wall Street, often dreamed up by art or creative writing majors
I have no idea what he's talking about here. In my experience, publicly-traded businesses (which would be the ones who make a commitment to shareholders
, not Wall Street
) don't usually have art or creative writing majors making major business commitments; the more likely case is that those positions are staffed with MBAs. Privately held businesses, on the other hand, tend to be beholden to venture capital firms, whose presence can become far more unwelcome and meddlesome than that of your typical shareholder.
We all know that there must be better way, right? DevOps is the proof that it’s possible to break the core, chronic conflict, so we can deliver a fast flow of features without causing chaos and disruption to the production environment.
YOU NEED MANAGEMENT BUY-IN. YOU NEED THE SUPPORT OF THE PEOPLE WHO CONTROL THE MONEY TO HIRE THE RIGHT PEOPLE AND PURCHASE THE RIGHT EQUIPMENT, THE PEOPLE WHO SET THE COMPANY'S EXTERNAL AND INTERNAL STRATEGIES, THE PEOPLE WHO HAVE YOUR BACK WHEN YOU TELL THE COMPANY, "THIS WILL HURT, BUT IN 12 MONTHS, WE WILL NO LONGER LIMP."
Painting DevOps as a panacæa does all of us who believe in DevOps and work in DevOps a tremendous disservice. To unfuck a fucked company, you need to fix the culture. This is why Netflix is a shining example — not because of their use of DevOps, but because their culture enabled them to use their talent in a massively constructive and creative fashion.
"Before you can solve a complex problem, you must first have empathy for the other stakeholders."
Before you can solve a complex problem, you must first understand
it. One of the important factors in understanding it is empathizing with the other parties. It is not, however, the first
thing to do.
Perhaps i'm being too harsh on what's less of a thoughtful article and more a plug for his book. But i do not find the appeal in being sold on DevOps by starting off with an elaborate strawman.
Sat, Dec. 8th, 2012 12:53 am (UTC)
Consider posting this in another place for comment?
Sat, Dec. 8th, 2012 12:58 am (UTC)
If you mean The Other Place, i'm currently bereft of access. Feel free to post a link there.
Sat, Dec. 8th, 2012 02:40 am (UTC)
Do you want access?
Sat, Dec. 8th, 2012 02:44 am (UTC)
Sat, Dec. 8th, 2012 02:48 am (UTC)
Give me an IP, and I can give you access on usenet.killfile.org.
Sat, Dec. 8th, 2012 03:03 am (UTC)
ennui.org has address 126.96.36.199
Sat, Dec. 8th, 2012 03:30 am (UTC)
Sat, Dec. 8th, 2012 03:32 am (UTC)
200 usenet.killfile.org InterNetNews NNRP server INN 2.5.2 ready (posting ok)
Sat, Dec. 8th, 2012 04:08 am (UTC)
(Is sidehack still connected to that place? I think I still have a shell account on that...)
Sat, Dec. 8th, 2012 05:53 pm (UTC)
Sat, Dec. 8th, 2012 02:11 pm (UTC)
*happy nostalgia for that class B*
Sat, Dec. 8th, 2012 04:07 am (UTC)
Engineers have an irrational dislike of humanities folks, when really we should all be ganging up against the business types.
Sat, Dec. 8th, 2012 05:27 am (UTC)
Developers wear grey. They work much harder than we do, because they're so frightfully clever. I'm really awfully glad i'm a sysadmin, because i don't work so hard. And then we are much better than the product managers and business people. PMs are stupid. They all wear green, and businesspeople wear khaki. Oh no, i don't want to play with businesspeople. And executives are still worse. They're too stupid to be able to read or write. Besides they wear black, which is such a beastly colour. I'm so glad i'm a sysadmin.
Sat, Dec. 8th, 2012 05:48 am (UTC)
I didn't get *any* soma this week.
Sat, Dec. 8th, 2012 04:40 pm (UTC)
jwgh: something nobody has ever said
I'm so glad I'm in tech support.
Sat, Dec. 8th, 2012 06:37 pm (UTC)
freelikebeer: I'm surprised ...
... that there can be places other than SuperParochialDefenseLand where devs don't have a deep, expressed interest in ops.
Sat, Dec. 8th, 2012 11:02 pm (UTC)
From what i can tell, it's a significant cultural disconnect between people who just want to work on the next cool thing and people who don't want anything to break. The cultures need to work to bridge the divide.
Sat, Dec. 8th, 2012 08:07 am (UTC)
Yes, you absolutely need buy-in at all (or at least all relevant) levels. Otherwise the whole thing goes horribly wrong and now you're worse off than you were when you started. And to get that the organisational culture needs to not be hostile to change.
You also need an environment where devops is even useful to start with. If what you have is a collection of what are effectively boutique systems all configured differently in significant ways for good reasons, then adding another layer of abstraction will not help
. There are sizable firms doing Serious Business(TM) for whom this is true as the world is not all web services companies.
In a software company it can be seen as a political power-grab by the development department, of course. That never ends well.
Sat, Dec. 8th, 2012 04:08 pm (UTC)
If it's done wrong, sure (and how often does Dev need to do a political power-grab? they're usually already there). If it's done right, it's a way for Ops to prove their value to everyone, instead of just being the people who spend money and fix the site when it's broken.
Sat, Dec. 8th, 2012 06:13 pm (UTC)
Clearly all companies should shrink down to about a half dozen people. That way, buy-in and consensus can be blocked more efficiently.
Sat, Dec. 8th, 2012 11:00 pm (UTC)
You misspelled, "shrink down to the size where we can drown it in a bathtub."
Sat, Dec. 8th, 2012 06:22 pm (UTC)
As far as I can tell, "devops" is a word for the realisation that you need sysadmins who can do more than the generic Windows progress-bar watcher. This news has disconcerted management since before Windows existed.
Sat, Dec. 8th, 2012 09:40 pm (UTC)
I left this open in a tab, intending to come back later and say something witty. But now that I've thought about it a little more, I'm forced to admit that the job I've had for the last four years has been so soul-destroyingly corrosive that I don't think I have anything useful to add. I can, however, write a pretty weighty process management book focusing on "How Not To Do It", because I have been living in a ten-billion-dollar example of it for so long.
The single biggest problem my current project faces is lack of adult supervision, and the reason it's all gone horribly pear-shaped is precisely because "before you can solve a complex problem, you must first understand
it", and no one here -- not dev, not ops, not management -- really does.