michael@??? probably said:
> I wondered if I should mention that it has been discussed before. ;-)
Many times :)
> Extra memory is a point, but did you actually see the context switch
> problem? If so, I would be interested in details.
I had a machine that seemed to be suffering from too many context
switches and calmed down a lot when I disabled the local nameserver,
I didn't have time to track it down alas.
If the machine is having trouble with the number of concurrent
deliveries then context switching to the nameserver to resolve all of
them can make things far worse. Hopefully a multiprocessor machine of
a decent arch shouldn't have this problem as badly.
> Then again, if the remote BIND goes down, you have a problem. This is
Thats why I suggest forward first; not forward only.
] forward
]
] This option is only meaningful if the forwarders list is not empty. A
] value of first, the default, causes the server to query the forwarders
] first, and if that doesn't answer the question the server will then
] look for the answer itself. If only is specified, the server will only
] query the forwarders.
If the forwarding nameserver goes down things won't break, they'll
just delay a little ... and your pager should be going off because
your monitoring system has detected the forwarding nameserver is down
so you fix it ;)
> a particularly religious matter. I guess we can agree that everybody
> should think about it and then decide. :)
Yep. It does seem to differ across machine arch, for example. I just
try to get people to think about it and test it, not blindly follow
advice one way or the other.
P.
--
pir pir@??? pir@???
