notify timed out/exceeded

John Wobus jw354 at cornell.edu
Thu Jul 28 15:24:00 UTC 2005


When I had this problem, it turned out that a buffer for incoming
UDP packets was filling and packets were being lost.  I found two
different ways to fix it:
-It went away when I switched to a newer/faster machine
-It went away when I changed an OS parameter to make processes
use a larger buffer for incoming UDP packets.

John


On Jul 28, 2005, at 8:01 AM, Gamer wrote:

> Dear list,
> im still evaluating the performance of multiple slaves with a Master
> Nameserver (zone transfers, notifies etc.). In order to simulate
> traffic with many slaves, I assigned ~10 virtual ip addresses to a
> single interface. With "listen on {..}" and "transfer source {..}" I
> attached the named processes to the virtual interfaces. Now I get tons
> of :
>
> Jul 28 13:17:16.414 notify: debug 1: zone test40/IN: notify to
> 10.65.95.236#5353 failed: timed out
> Jul 28 13:17:16.414 notify: debug 1: zone hugo40/IN: notify to
> 10.65.95.251#53: retries exceeded
>
> messages in my log files...
>
> The Master sent the notifies properly (log+tcpdump); the
> slave(s)received the notifies (tcpdump) but there is no log entry!
> I`m quite sure the packages don`t get lost in the network.
> Any idea about that? Is it bind (9.3.1) or maybe the virtual
> interfaces?
> Thanks in advance!
>
>



More information about the bind-users mailing list