BerkeleyDB locking subsystem

[Peter Corlett]

On 1 Aug 2006, at 12:44, Dominic Mitchell wrote:
[...]
> Of course, if the process is stuck deep inside the kernel, it's  
> possible
> for it to have the appearance of not being killed by a SIGKILL.  If  
> that
> happens, file a bug about your O/S.

That'd be the infamous uninterruptible sleep (which shows up as  
status D). It also manifests itself on hardware failure and when the  
box is thrashing.

I've been cussing a high-traffic box today that had about 30 CGI  
scripts in D state. Note I didn't say FastCGI or mod_perl. It also  
doesn't have "firefighter" or "sysadmin" in my job description...