Other than KILL?

Joseph S D Yao jsdy at cospo.osis.gov
Tue May 16 19:51:43 UTC 2000


On Tue, May 16, 2000 at 03:36:49PM -0400, Kevin Darcy wrote:
...
> a "hard" kill can fail if the process is executing a buggy/misbehaved kernel
> call, but I've never seen an unkillable named).

NOT buggy/misbehaved.  Possibly disk controller error.  This is a
feature, in some ways - it provides for atomic disk writes, if the
hardware is working.

> One caveat: on a really, *REALLY* busy system, I suppose it's theoretically
> possible that you might end up killing the wrong process with this technique
> (if the PID's wrapped in those few seconds)....

You might try testing whether the PID file still exists AND has the
same number in it.

...
> Cinense, Mark wrote:
...
> >	..  Is there any way of providing nameservice that produces .9995
> > uptime.  The customer wants to have nameservice with these stats 24x7.  Is
> > this possible?

This is what slave servers are for.  Distribute them on different
networks and geographically so that no two are affected by power or
network outages at the same time, and you increase your MTBF enormously.

It depends on how much they are willing to pay, to get near-24x7.

-- 
Joe Yao				jsdy at cospo.osis.gov - Joseph S. D. Yao
COSPO/OSIS Computer Support					EMT-B
-----------------------------------------------------------------------
This message is not an official statement of COSPO policies.



More information about the bind-users mailing list