Bind 9.2.1 - socket.c:1100 unexpected error

Mark_Andrews at isc.org Mark_Andrews at isc.org
Wed Jun 26 00:20:58 UTC 2002


> 
> Hi,
> 
> I am running bind-9.2.1-0.7x on a RH Linux 7.2 DNS server.
> Each night a lot of dns lookups are performed by Webtrends on another
> machine to process firewall logs.
> And since I upgraded to 9.2.1, I am getting a lot of those error messages in
> the logs :
> 
> Jun 25 04:58:44 XXXXX named[29123]: socket.c:1100: unexpected error:
> Jun 25 04:58:44 XXXXX named[29123]: internal_send: XXX.XXX.XXX.XXX#29722:
> Connection reset by peer
> Jun 25 04:58:44 XXXXX named[29123]: client XXX.XXX.XXX.XXX#29722: error
> sending response: connection reset
> 
> This is certainly due to an overload of the bind process but why this did
> not occur using bind 9.1.3-4 ?
> 
> I tried to change named.conf and increase recursive clients setting to :
> ...
> recursive-clients 1500;
> ...
> 
> But this did not change nothing to the problem.

	It wouldn't as you are not exhausting the recursive clients.

	What is happening is that the client is closing the connection
	with a reset rather than a shutdown.  Normally if the client
	performs a graceful shutdown you will get EPIPE not ECONNRESET.

	Mark

Index: bind9/lib/isc/unix/socket.c
diff -u bind9/lib/isc/unix/socket.c:1.227 bind9/lib/isc/unix/socket.c:1.228
--- bind9/lib/isc/unix/socket.c:1.227	Sun Jun  2 23:31:16 2002
+++ bind9/lib/isc/unix/socket.c	Tue Jun 25 17:12:54 2002
@@ -1092,6 +1092,7 @@
 		ALWAYS_HARD(ENOBUFS, ISC_R_NORESOURCES);
 		ALWAYS_HARD(EPERM, ISC_R_HOSTUNREACH);
 		ALWAYS_HARD(EPIPE, ISC_R_NOTCONNECTED);
+		ALWAYS_HARD(ECONNRESET, ISC_R_CONNECTIONRESET);
 
 #undef SOFT_OR_HARD
 #undef ALWAYS_HARD

> Need some advice.
> 
> Thank you.
> 
> Anthony
> 
> 
> 
--
Mark Andrews, Internet Software Consortium
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark.Andrews at isc.org


More information about the bind-users mailing list