Date: Thu, 5 Mar 1998 23:04:10 -0500 (EST) From: Robert Watson <robert@cyrus.watson.org> To: Leif Neland <leifn@image.dk> Cc: freebsd-hackers@FreeBSD.ORG Subject: Re: Cluster? Message-ID: <Pine.BSF.3.96.980305224859.241B-100000@cyrus.watson.org> In-Reply-To: <520_9803060120@swimsuit.swimsuit.roskildebc.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On 5 Mar 1998, Leif Neland wrote:
> At 05 Mar 98 17:54:20 Tom (2:234/49.99) wrote to All regarding Re: Cluster?
> in area "freebsd-hacker"
>
> >> address spaces of all the others. I agree that DNS has fault tolerant
> >> capabilities but I wouldn't call them excellent, its annoying to have to
> >> wait for requests to the primary to time out before the resolver
> >> library starts hitting on the secondaries.
>
> T> Speak for your own resolver. I can't even tell if a primary
> T> fails.
>
> In one case I know, the first timeout is 5 seconds, before trying the
> secondary.
> The next timeout is 10, then 20 and finally 40 seconds. Your mileage may vary.
However, I believe that the standard resolver library will store which
servers respond and which don't, and favor servers that have responded in
the past. This is certainly true if it receives a connection refused icmp
packet, but I'm not sure about lack of response:
(res_send.c):
/*
* On a 4.3BSD+ machine (client and server,
* actually), sending to a nameserver datagram
* port with no nameserver will cause an
* ICMP port unreachable message to be returned.
* If our datagram socket is "connected" to the
* server, we get an ECONNREFUSED error on the
next
* socket operation, and select returns if the
* error message is received. We can thus detect
* the absence of a nameserver without timing out.
* If we have sent queries to at least two
servers,
* however, we don't want to remain connected,
* as we wish to receive answers from the first
* server to respond.
*/
The badns flag for that particular name server is set so that we avoid it
in the future. A quick scan of the code did not turn up a badns bit
toggle on timeout, as I think it is assumed that server disappearance may
be transient, or the result of high load? I don't believe code to shift
down precedence of a server exists in the v4.9.x BIND code, but it would
certainly be a useful addition.
I have not checked the Bind 8 code, however.
Robert N Watson
Carnegie Mellon University http://www.cmu.edu/
SafePort Network Services http://www.safeport.com/
robert@fledge.watson.org http://www.watson.org/~robert/
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.96.980305224859.241B-100000>