Distributing DNS servers

Barry Margolin barmar at bbnplanet.com
Tue Aug 31 14:44:59 UTC 1999


In article <Pine.BSF.4.01.9908301932580.5597-100000 at phoenix.aye.net>,
Barrett Richardson  <barrett at phoenix.aye.net> wrote:
>> likely that named would die or hang than that a Sun workstation would
>> crash; since we have to implement monitoring for that and procedures to
>> remove the static route manually in such cases, it wouldn't be worth the
>> extra complexity of running a routing protocol on the server just to handle
>> one specific case.  Some day we might change this; for instance, if the
>> named monitoring could automatically reconfigure gated to stop advertising
>> the route, it would be reasonable to run gated.
>
>What I had in mind was to run gated, and have a code on the same box
>that endlessly sits in a loop sending queries at regular intervals. If
>named does not respond in a specified time interval (open to suggestions
>on the time interval), then just assume it is sick and send a SIGTERM
>to the pid in /var/run/gated.pid. With BGP the route removal should
>be immediate, I'll have to test what OSPF does when a gated neighbor
>dies.

This is pretty much what we're thinking we might do in the future.
Currently we're doing remote monitoring, so it's not as easy to reconfigure
the machine automatically when it notices a failure.

As for OSPF detecting dead neighbors, it won't be as quick as BGP.  BGP
uses TCP connections between peers, and killing gated will cause the
connection to close, which the peer can detect immediately.  OSPF uses
periodic multicasts, so a dead peer is only noticed when you miss several
announcements in a row.

>I haven't scoured through the bind distribution much but there has to
>be some code there I can cannibalize to send queries.

Why not just use dig in a shell or perl script?

-- 
Barry Margolin, barmar at bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.


More information about the bind-users mailing list