[lug] Memory Leak Detection

[Nate Duehr]

To test for a memory leak, just chart the memory use over time during a "quiet" time for the machine. 

If real memory use keeps going up (after the initial Linux greediness to cache as much of the filesystem as possible -- don't look at numbers that include filesystem cache!) then it's "leaking". 

As far as the consultant goes... You've probably seen the poster/picture from demotivators.com -- "Consulting: If you're not part of the solution, there's good money in prolonging the problem!"

Heh. Heh. 

I just had the joy of running into a counter rollover bug at work last week... Timer that runs at a speed that takes 497.1 days to rollover, then... Crunch.  A couple million dollars worth of gear goes out to lunch for an "unknown" reason. Wheeeee... 

Made for an un-fun day.  Gotta love it when you lose a Saturday to someone else's bug they wrote years ago... And yes, the fix was released at least a year ago, the customer just didn't/couldnt upgrade. 

There should be warning lights and sirens with stories like that one posted somewhere on the "road to sysadmin land" for people actually considering becoming one!  Hahah...

The only good news about being a sysadmin is that I haven't met a coder yet who could release bug-free code.  

That's a strange commentary on both job security and the sad state of discipline in our industry.

--
Nate Duehr, nate at natetech.com - WY0X