Slashdot Mirror
Should Developers Have Access To Production?
WHiTe VaMPiRe writes "Kyle Brandt recently wrote an editorial exploring the implications of providing developers access to the production servers of a Web site. He explores the risk introduced by providing higher level access as well as potential compromise solutions."
Whenever an error occurs that I can't replicate in a dev environment, I'm always SO tempted to hop into prod and start adding in some output statements.
Yeah, it's probably a good thing I don't have access to prod.
LOL! No.
No. It just encourages sloppy development practices.
Would you want to drive over a bridge that wasn't actually designed and engineered, but rather they just piled some stuff up and will fix it if it collapses? Or have a surgeon chopping you open with the idea that they'll figure it out as they go? So why would we want developers to work with the expectation that they get to intervene at the last instant to resolve their failures?
It is my experience that giving development access to production gives you a production environment that looks like it has been vandalized. Although meaning well and trying to make the best application as possible; they need their own development lab, and their own staging / production lab.
Everyone agrees that developers should never have access to production...Unless they're the developer, in which case it's different.
Its a good practice to keep them separated, but in the end its just a pissing contest. The server admins don't want some filthy dev messing with their stuff, and I can appreciate that.
However, admins often lack appreciation of some dev-specific issues, and their ignorance can lead to problems down the line.
In the end, its the best practice to have everyone work together sensibly, than throw down inflexible rules that cause more trouble than they prevent.
ad logicam Claiming a proposition is false because it was presented as the conclusion of a fallacious argument.
So, yes.
This is why there is a change control process, and a testing environment.
If you're doing it wrong, you're asking for trouble.
The price is always right if someone else is paying.
There's no correct answer to this question. It depends on the size of the organization and the nature of the system. I've worked in different companies that have been on either sides of where I thought the line should be. The line is drawn in a very different place for a 20 employee company than where it is in a 20,000 employee company.
I think it's helpful in analyzing real-world data and getting an idea about real system loads, testing issues to see if they are in the wild today, etc. For a good developer, it makes life much easier.
In a very healthy development ecosystem all this data is replicated and there is never any need for a developer to touch prod. In the development ecosystems that exist in the real world though, most are very unhealthy, frustrated by ham-fisted security, process flaws, red-tape, inconsistency, and incompetence ranging from scattered to mostly cloudy.
The answer is, do you have the class of developer that knows what not to do and desires to play nice, or do you have the usual.
As a developer I can tell you that it's impossible to test programs properly and thoroughly without access to production data. However, developers should NOT be granted access to production logins/sites - production data should be copied into development work areas so that developers have an appropriate "sandbox" in which to work/test.
I don't think he meant malicious backdoors. I read that as backdoors to allow debugging/etc.
Biggest issue my cow-orkers and I have is that the sysadmin *claims* that the dev box and production box have the same packages, configuration, etc. but in reality, they don't. Most often we find out when we ask for production stuff to be copied over to the dev site to test errors, etc. and just loading it - which works on the live site - generates errors on the dev site.
Don't blame me, I voted for Kodos
If you are a small software shop then I can see reasons for allowing your small technical staff to have access to production. It's all well and good saying that only the admin of that server should have access and there's a full rollout procedure in place to be followed only on certain days, certain times; but even when I've seen that sort of structure in place there are times when it's useful for the developers to have access to production. Nothing is perfect and we'd all love to have multitude's of staging servers, replicating the typical load and uses of production but for a hell of a lot of (non critical I'd add) systems that just doesn't happen.
There simply is no one rule fits all. Sometimes I wish we had extremely rigorous rules & regulations in place - I'd probably get to go home a hell of a lot earlier. I'm not suggesting you start chucking exceptions all over your checkout code on live but I think you should asses your own situation (and staff for that matter).
jaymz
It's not necessarily a case of the admins versus the developers, its more of practicing good data governance.
Our developers used to have direct access to all of the production databases. This was bad enough, but because of this the organization permitted them to directly "clean up" databases (meaning they wrote to tables directly), we had data that was being changed without the ability to really know who did it. The DBAs hated it and the developers were extremely uncomfortable doing it but it happened anyway. We eventually had a real process audit and the auditors had a field day.
Needless to say we changed. I hope.
Even when the suggestion of "would you like root on this internal box?" was put to me, my answer was always "No". I write code. Others test it. Admins deploy it.
People specialize for a reason. If you want half-assed administration, give root to a developer. If you want half-assed code, let admins write software. If you want half-assed testing, have admins and/or developers do it.
For all intensive purposes, "whom" is no longer a word. That begs the question, "who cares"?
If I found out a developer changed something in a system I tested without it going through the proper process...
Let's just say I would be very interested to hear why they shouldn't go back and rerun everything again on THEIR dime. (at the very least) In fact, we DID do just that to someone who let a revision slip into their UUT because a developer felt it would fix something and make it perform better.
It wasn't too expensive of a mistake, just $250,000 to rerun that portion of the test. Although that was just the physical cost of performing the test. I don't even want to know how much it cost in labor especially considering it was a 22 day test.
Even if the change was removed, how do I know that without physically verifying checksums (do I even trust it anymore since their CM process is obviously flawed)
Out of modpoints but really liked a post? 1BDkF6TtmmeZ3yqXbz9yhdYVqRYnwFoXDj
Comment removed based on user account deletion
The more developers work in production, the more they can ONLY work in production.
I'm all for read access (the more eyeballs the better), but actual access to change anything is a train wreck. The devs will forget to check the changes in to the source repo, or they'll check them in differently (bad copy/paste), or they'll check them into the wrong branch/tag. Regardless the next release that goes out silently adds the bug back into production.
And if developers think it's difficult to fully clone a prod environment configuration into dev now, wait until they try to do it after developers have been hacking on it directly for a while.
Pretty soon every release is a train wreck requiring tons of post-release tweaking and hammering to get it in place. Every release is a stressful mess as you're all crossing your fingers because you really have no idea what you are actually changing and no way to find out.
Just don't do it. Hire a good build engineer/release manager/software configuration manager that can sort out, automate, and track environment management well enough that yes, you can reliably clone an accurate representation of production in a matter of minutes. He'll cost you about as much as a good sr developer, but the savings across the board will easily dwarf his salary.
My
Long answer, it depends on the situation.
I (and a couple of my coworkers) have access to production servers, but we don't develop on prod. End of story. We have other devs who do not have access to prod. Dev is for dev, prod is for prod and don't let anyone without the discipline to keep that rule have access to prod.
LK
"Hi. This is my friend, Jack Shit, and you don't know him." - Lord Kano
Hell. No.
I'm a developer as well as a sysadmin and I NEVER tweak anything in production and I have full access to it.
I have an exact copy of my production environment for development and I do all my tweaking/test deployments there.
In fact nothing gets deployed to production until everything has been checked in development.
My previous job had dev/qa/prod environments where the devs had full access to development and it was so bad that we had to virtualize it for them just so we could revert back to a pristine snapshot whenever they jacked up the dev server.
I don't know where you work, but when dealing with a few hundred or thousand tps per server, logging can account for a good chunk of latency.
Realities just a bunch of bits.