Do I dare declare? (Approaching the magical 770 minutes of processing time...)
Written by
on
in
Snaking.
The demo instance has been crashing with memory errors after about 770 minutes of processor time, or about 12 hours of wall-clock time. As of now, we're up to 767 minutes, with only 153MB of memory used.
So often these leaks have tormented me; pretending to be gone, only to reappear ever-so-much slower than before. But 153MB, how could anything hit a memory error with so much headroom? Do I declare the leak dead? Can I trust the reports of its demise enough to start working elsewhere in the system?
I dread that as soon as I do; as soon as I move on to the next item on my todo list (oh, how I long for that next item, I yearn for it... I'm afraid I may have conceived an unhealthy lust after it!) the leak will reappear to taunt me.
It's a vicious thing, full of cunning and hatred, wanting nothing so much as to crush my fragile spirit between the edge of its silicon memory and the hard, hot surface of its swap drive.
LLLLEEEEEAAAAAKKKKK!!!!!!
Comments
Comments are closed.
Pingbacks
Pingbacks are closed.
Mike Fletcher on 12/10/2004 4:44 p.m. #
Blah. Still a leak, about 3.5MB/hour. Or around 13 days before memory exhaustion... better than 12 hours, but dang that's annoying. Bleep.
Mike Fletcher on 12/11/2004 1:12 p.m. #
With a longer sampling rate, it's actually 6.5 MB/hr :( . Glyph tells me, btw, that there's a __del__ method on Deferred objects, which explains why the gc wasn't deleting them as expected.