Problems with interfaces going down

[Mark Andrews]

> On 13 Feb 2021, at 10:33, bindusers at prograde.net wrote:
> 
> Greetings,
> 
> I’ve been fighting a two-fold problem with named (bind 9.16.11) running on macOS.
> 
> 1: If an ethernet interface being listened to drops link, named immediately stops listening to it:
> 
> 12-Feb-2021 17:33:19.326 no longer listening on 192.168.88.220#53
> 
> and
> 
> 2: when link returns I get 2 tries to reestablish listening:
> 
> 12-Feb-2021 17:33:39.458 listening on IPv4 interface en1, 192.168.88.220#53
> 12-Feb-2021 17:33:39.463 creating IPv4 interface en1 failed; interface ignored
> 12-Feb-2021 17:33:41.946 listening on IPv4 interface en1, 192.168.88.220#53
> 12-Feb-2021 17:33:41.951 creating IPv4 interface en1 failed; interface ignored
> 
> which both fail because named is no longer running as root.
> 
> --------------
> 
> Where I’m confused is that this ISC KB article:
> 
> https://kb.isc.org/docs/aa-00420
> 
> seems to imply that the "no longer listening" event is due to a periodic interface scan finding the interface "unavailable".
> 
> That doesn’t fit my observations since it happens as soon as link is lost. If some minutes-long periodic scan were needed to detect the interface being down it would take, on average, half of that period to happen. It does not.
> 
> Further, I tried what the KB article advised by adding the option:
> 
> 	interface-interval 0;
> 
> That does seem to stop the periodic scan (since my log is no longer filled with errors) but the “no longer listening” event still occurs right when the interface drops.

``automatic-interface-scan``
   If ``yes`` and supported by the operating system, this automatically rescans
   network interfaces when the interface addresses are added or removed.  The
   default is ``yes``.  This configuration option does not affect the time-based
   ``interface-interval`` option; it is recommended to set the time-based
   ``interface-interval`` to 0 when the operator confirms that automatic
   interface scanning is supported by the operating system.

   The ``automatic-interface-scan`` implementation uses routing sockets for the
   network interface discovery; therefore, the operating system must
   support the routing sockets for this feature to work.

> --------------
> 
> Is it not possible to have named drop to a non-root user (via -u) but still recover from (or ride through) a momentary ethernet link loss?

If you allow euid switching then you may as well not run with -u as the whole point of -u is
to be a safeguard in the case where there is a program flaw that allows arbitrary execution
to occur.

> Having the server stop working due to a switch I have no control over burping is very suboptimal.
> 
> Thanks for any ideas.
> 
> -Mike
> 
> _______________________________________________
> Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe from this list
> 
> ISC funds the development of this software with paid support subscriptions. Contact us at https://www.isc.org/contact/ for more information.
> 
> 
> bind-users mailing list
> bind-users at lists.isc.org
> https://lists.isc.org/mailman/listinfo/bind-users

-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742              INTERNET: marka at isc.org