Monday, October 25, 2010

Why I think Devops is stupid

http://www.jedi.be/blog/2010/02/12/what-is-this-devops-thing-anyway/

First of all, this isn't a "movement." People have been trying for years to get quality sysadmins who are also competent programmers. I still believe that except for a few rare cases, these people do not exist. And they shouldn't: clearly, something is wrong.

If I told you I spend all of my time both becoming the best sysadmin I can be, and becoming the best programmer I can be, would you believe me? If so, I have a bridge to sell you. The fact is that when i'm a sysadmin I really don't program much at all. I spend my day at work fighting fires and performing odd jobs and when I get home the last thing I want to do is get back to the computer. And at work, if I spent most of my time researching new development trends and writing new tools in experimental languages, how much real sysadmin work am I doing? No, the truth is I wouldn't have enough time in the day to be both a full-time sysadmin and a full-time programmer. I can only do one job at a time.

"the Devops movement is characterized by people with a multidisciplinary skill set - people who are comfortable with infrastructure and configuration, but also happy to roll up their sleeves, write tests, debug, and ship features."

Sorry. I have a job. I don't want to have to do the developers' jobs too. I'm upgrading the Oracle cluster to RAC and being woken up at 3 AM because some bug somewhere deep in the site caused pages to load all funky, and i'm trying to figure out who committed the flaw and get them to revert it. Even if I wanted to, i'm a sysadmin; i'm not familiar with the developers' codebase, and sometimes not even the language they're writing it in. How the hell can you expect me to realistically debug it in real time? And writing tests? Really, you want me to write the developers' unit tests?

Don't get me wrong. I am fully in support of the general idea of better communication between groups and sysadmins working with developers, DBAs, QA, neteng, etc to build a better product. I think it'd be insane for any group to go about making any major changes without consulting every other group and working out any potentially negative ramifications. But this doesn't mean each group has to know how to do each other group's job. Communication is the key word here, not cross-pollination.

There are lots of technical issues that come up in the building of any product. To make it work as well as possible, there's lots of different problems which have to be accounted for. The problems cited in the above post - 'fear of change,' 'risky deployments,' 'it works on my machine,' 'siloization' - all require planning and cooperation to resolve. But this is basic stuff, to me. You don't need to be a DevOp to realize you're going to need your devs to have the same baseline system for testing their apps as your production system (sometimes more than one). The apps have to be developed in a way that allows for a smooth upgrade in the future. And you need a competent deployment and reversion system with change approval/code review and reporting.

These issues are not solved by simply having a 'DevOp', whose responsibility is not only their own systems but apparently the total management and architecting of the whole process of development of a product and delivering it working flawlessly. To properly deal with these issues you need many things. You need really strong management to keep teams working together and to help them communicate. You need some kind of manager or architect position who can keep track of how everything works and juggle the issues before they become serious problems. You need people who are really good at doing their job and get them to ask for help.

Nobody's job is simple. But creating some new position to supposedly solve all these issues by being super-human techno-gods? Even if you could get these godly Devops people in every corporation, there's no promise that they can even get past the politics inherent to each group to make everything work as harmoniously as the post describes. There is no magic bullet. No movement will make everything alright. The world is harsh and complex, and a DevOp isn't going to save it.

1 comment:

  1. The most dangerous thing about DevOps is that it seems to be a thinly veiled attempt to short-circuit the traditional development cycle (write it, test it, stage it, change management, production) and to bypass the operations people who, rightly, do not want bad code running in production. The result is unstable, insecure code being recklessly promoted from, as we say, dev to prod. If it breaks, it is patched willy nilly. Ever more frightening is the number of potentially exploitable PHP and memcache running free and clear on public IP space in "the cloud" because so many are too dumb, lazy or poor to pay for a sysadmin.

    ReplyDelete