[Swan] Pluto consumes all available memory

Paul Wouters paul at nohats.ca
Thu Apr 30 23:38:25 EEST 2015


On Thu, 30 Apr 2015, Will Roberts wrote:

> Here's  what I had after about 36 hours.
>
> Apr 30 20:03:48 albuquerque pluto[30192]: leak: 4 * struct event in 
> event_schedule(), item size: 32
> Apr 30 20:03:48 albuquerque pluto[30192]: leak: pluto crypto helpers, item 
> size: 64
> Apr 30 20:03:48 albuquerque pluto[30192]: leak detective found 5 leaks, total 
> size 96
>
> That doesn't seem like a whole lot to me, and making that 20 times worse (for 
> 20x the length of time) doesn't seem much worse either. I'll deploy this to 
> all of my servers and report back in a while.

So possibly, the leak only happens in some failure case that keeps
repeating. I'd suggest to leave it running longer. Look at the pluto
memory size, and only when it gets really big do the restart and check
for leaks.

Paul


More information about the Swan mailing list