Re: posible bug in daemontools

On Sun, Mar 21, 2010 at 10:55:40PM -0600, Andy Bradford wrote:

> While  this certainly  might be  a  bug in  daemontools, it's  debatable
> whether or not a fix is a good  thing. Whenever I have found a system in
> which  the kernel  started  randomly killing  off  processes, it  almost
> always  (if indeed  not  always)  means that  the  system  needed to  be
> rebooted.

I agree with your last point about rebooting after the kernel was forced to
kill processes, but that's not what happened: the system was low on memory
so fork temporarily returned an error. *No* processes were killed, only fork
failed. Daemontools never retried the failed fork, and the service remained
down (in "want up" state).

-- 
Jurjen Oskam

Savage's Law of Expediency:
        You want it bad, you'll get it bad.