If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Naked Security)   Should computer programs be intentionally terminated every so often in order to inspire designers and coders to make them more robust? Some call this practice 'shut up and reboot', I call it 'pour beer on the server'   (nakedsecurity.sophos.com) divider line 28
    More: Interesting, web app, computer programs, vacation time  
•       •       •

1629 clicks; posted to Geek » on 29 Jul 2013 at 11:59 AM (1 year ago)   |  Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



28 Comments   (+0 »)
   
View Voting Results: Smartest and Funniest
 
2013-07-29 12:25:24 PM  
Should the "Too ling; didn't read" theory be applied to other situations?

That article makes me think there needs to be a short acronym similar to "tl;dr" expressing like
"incoherent rambling; have forgotten everything 10 seconds later"
or
"doesn't make point, waste of time"

/otherwise I'll have to fall back to "your block sucks" and point out that this piece leads to some post at ars technica which was talking about a question on stackexchange
 
2013-07-29 12:25:31 PM  
In testing environments? Sure. In productions, absolutely not.
 
2013-07-29 12:41:17 PM  
I will consult with my manager regarding the creation of a randomly timed task on our production webserver to terminate IIS, forcing a manual restart as a means of taking the advice of the article.
 
2013-07-29 12:42:24 PM  
My manager said "no".
 
2013-07-29 12:50:42 PM  
Wow, that writer.
 
2013-07-29 12:53:00 PM  
Isn't there some cloud provider that basically has an on-going demon roaming around their network purposely screwing with it to test the resilience?  I think its a noble concept for some situations, unimportant crap like clouds, your web browser or maybe a few desktop apps (no auto-save configured in word?  hahahaha sucks to be you!).  It'd teach people the value in backups and get some good habits formed.

Meanwhile, for everything else:  No, that's probably one of the stupidest things I've ever heard.  I hope this guy is on a plane who's software was designed with his omniscience vision in mind.

PS:  Your blag sucks.
 
2013-07-29 12:59:52 PM  
If it wouldn't take them 10 years to get it back to where it is I sometimes wish Adobe would do that with it's products.  So many legacy holdovers that bog it down.  But there was an article on here not to long ago talking about how so many people want that old shiat that you'd piss off too many users in your attempt to overhaul it.  Also it would take 10 years.
 
2013-07-29 01:07:25 PM  
How about no?  Fix your goddamn memory leaks
 
2013-07-29 01:16:28 PM  
Dimensio
I will consult with my manager regarding the creation of a randomly timed task on our production webserver to terminate IIS, forcing a manual restart as a means of taking the advice of the article.


Strangely enough, the one kind of program doing something similar that came to mind, were (web) servers like Apache which can terminate child processes and start new ones after the old child process has handled x number of requests or used x bytes of memory during its lifetime.

BumpInTheNight
Isn't there some cloud provider that basically has an on-going demon roaming around their network purposely screwing with it to test the resilience?


Probably Netflix because that was mentioned in the original post as the thing that got the poster thinking about whether that allegedly successful approach could make sense elsewhere, i.e "if it makes sense on the macro level (nodes in distributed systems) , could doing something similar at the micro level (a program's threads) have advantages?"
 
2013-07-29 01:26:28 PM  

dehehn: do that with it's products.


Could you terminate that extra apostrophe?
 
2013-07-29 03:11:48 PM  

MindStalker: In testing environments? Sure. In productions, absolutely not.


Netflix disagrees with you. Chaos Monkey Service will ruin your day if you are not confident to run it on production.
 
2013-07-29 03:13:20 PM  

The Slush: How about no?  Fix your goddamn memory leaks


It is about finding shiat code that should handle a fault cause they will happen and often with cloud computing.
 
2013-07-29 03:53:17 PM  
Maybe in a COTS environment, but anything mission critical should strive for as long an uptime as possible. If Netflix's "Chaos Monkey" fails, then someone reboots their video. If the ISS ECS system fails, there are bigger problems than an error message.
 
2013-07-29 04:20:12 PM  
If you're gonna do that, then it should at least start up a copy of itself and have some way of seamlessly switching over to use the new instance before closing down.
 
2013-07-29 05:16:20 PM  
It's a good design guideline to support dynamic unloading of anything that can be dynamically loaded. That's what he was saying, right?
 
2013-07-29 05:17:29 PM  

Gig103: Maybe in a COTS environment, but anything mission critical should strive for as long an uptime as possible. If Netflix's "Chaos Monkey" fails, then someone reboots their video. If the ISS ECS system fails, there are bigger problems than an error message.


fc06.deviantart.net
 
2013-07-29 06:16:11 PM  
If it did that with one of my programs out in the field, it could kill someone.
 
2013-07-29 06:21:41 PM  

meat0918: If it did that with one of my programs out in the field, it could kill someone.


This.

If you want to do that in THE TEST ENVIRONMENT, let's chat.
If you want to do that in PRODUCTION, please DIAF.
 
2013-07-29 08:11:07 PM  
No, because rewriting programs from scratch is one of the highest risk activities you can do.
 
2013-07-29 08:47:34 PM  
I'll plant that idea in my next DO-178 Software Development Plan and see if it survives the review process.

Maybe it'll get me one of those much-needed rests the article talked about.
 
2013-07-29 09:28:05 PM  
Writing crappy code that can't run for more than a day without faulting is pretty much SOP today.

Calling your failure intentional, and heralding as a feature?  Impressive.  Fast track to management for that one.
 
2013-07-29 09:37:11 PM  

Intrepid00: MindStalker: In testing environments? Sure. In productions, absolutely not.

Netflix disagrees with you. Chaos Monkey Service will ruin your day if you are not confident to run it on production.


If Netflix goes down in production it's hardly the end of the world, though. I would be a lot more hesitant to try this technique on a critical system.
 
2013-07-29 10:27:52 PM  

The Voice of Doom: Should the "Too ling; didn't read" theory be applied to other situations?

That article makes me think there needs to be a short acronym similar to "tl;dr" expressing like
"incoherent rambling; have forgotten everything 10 seconds later"
or
"doesn't make point, waste of time"

/otherwise I'll have to fall back to "your block sucks" and point out that this piece leads to some post at ars technica which was talking about a question on stackexchange


Try again sober
 
2013-07-29 11:34:19 PM  
I think I've figured out how to get my Unix admins to reboot servers. Tell them it's a resiliency test. That way I don't have to relive this for the nth time.

My Unix admins: "This critical service keeps hanging for no discernible reason. We have no clue why. It's the only thing on that box."
Me: "What's uptime on that box?"
My Unix admins: "NOOOOOO! We just need to restart the daemon when it hangs for no discernible reason."
Me: "Have you noticed that it's getting progressively shorter between hangs?"
My Unix admins: "NOOOOOO! There's nothing that doing a system restart will do that bouncing the daemon doesn't."
Me: "...and does 'bouncing the daemon' include a kill -9 every time? Humor me."
My Unix admins: :NOOOOOO! Maybe there's a patch in Perl that fixes a memory leak. Please, we don't have wives or girlfriends, our uptime is all we've got."
Me: "Give me sudo and I'll do it myself."
My Unix admins: "No, no, that's OK. Look! Restarted!"
Me: "Uptime: 735 days, 12 hours, 5 minutes 32.2345 seconds, you lying liars who lie. I'm pulling the plug."
My Unix admins: "NOOOOOOO! You Windows-loving corporate sellout, we deny you, we have no boss."
Me: "Three weeks, no hang. Theories?"
My Unix admins: "We have no clue, but it wasn't rebooting, because that doesn't work on Unix boxes."
 
2013-07-30 01:11:54 AM  
No.
 
2013-07-30 02:52:16 AM  
This tech exists, its called Vista.
This tech exists, its called XP.
This tech exists, its called ME.
This tech exists, its called "a mac".
This tech exists, its called "EA Games"
This tech exists, its called Chrome.
 
2013-07-30 06:43:22 AM  
Would be nice if computers in the workplace did that every few years so employees of misely owners wouldn't have to put up with painfully slow systems

/is thankful that his bosses are not at all miserly
 
2013-07-30 03:08:36 PM  
Absolutely not. If that's normal for production, that's what people will test to. Then you'll get the crappiest code known to man because, "eh, the process killer will clean it up anyway."

I could see it almost being useful in cloud though. Sort of a random load balancing... but I'm pretty sure there's cleaner ways to do that.
 
Displayed 28 of 28 comments

View Voting Results: Smartest and Funniest


This thread is closed to new comments.

Continue Farking
Submit a Link »






Report