How does it work the Secondary Server?

Balasubramanyam, Shivakumar sbalasub at qualcomm.com
Mon Apr 2 21:38:06 UTC 2007


Ricardo, ALL,

I have the exact same problem (case b.)

I used libresolv library and here is the behavior I found.

Actually, in case b, there are again two cases.
1. Machine is Up but named is down.
2. Machine is down.

In first case,
1. Using UDP.
	Every request goes is to primary/secondary in a round robin
fashion.
	Without lowest timeout of 1 sec, this is still very slow.
	I could not figure out how to tell the client library to stick
to one 	that 	last succeeded.

2. Using TCP.
	Much faster than UDP as it quickly gets a RESET packet for a SYN
request.
	TCP_STAYOPEN did not work in this case.I was expecting to use
the same 	connection for additional requests

Second case,

1. Using UDP,
	same behavior
2. Using TCP
	The SYN packets are not responded and tcp stack takes about 5
minutes to fail on the fist machine. SYN_ACK_RETRIES was set to 5. I got
to down to  less than 30 seconds  by setting the parameter to 2.
So if you are performing two queries, it takes 10 minutes to succeed. 

But when I use the "host" or "dig" application, it is really fast in all
the above scenarios. 

Any idea what I am doing wrong?

Thanks,
Shiva

>-----Original Message-----
>From: bind-users-bounce at isc.org [mailto:bind-users-bounce at isc.org] On
>Behalf Of Ricardo Martinez
>Sent: Monday, April 02, 2007 2:13 PM
>To: bind-users at isc.org
>Subject: How does it work the Secondary Server?
>
>Hello list.
>	I have a general question about how the seconday/slave server
works.
>Please correct me if i'm wrong.
>I have two DNS bind servers, one is the primary server and the other,
as
>you could guess, is the secondary/slave server.   My question is
related to
>the way the requests are procesed in
>
>a.- Normal Situation
>b.- Failure of the Primary Server
>c.- Failure of the Secondary Server
>
>Can someone explain to me how this work, i'm a little bit confused
>especially when a failure occurs.
>For example in normal situation the request from the root servers
arrives
>50% to my primary server and 50% to my secondary, isn't?  If so, what's
the
>idea of a secondary server if primary goes down? it will keep receiving
>requests even if the server is down?
>
>Hope that someone could explain to me this.
>Thanks in advance
>
>Regards,
>Ricardo.-
>



More information about the bind-users mailing list