bind-9.4.0-b3: internal_send: 0.0.0.1#53: Invalid argument and CPU locking

[Marco Schumann]

Hello,

bind-9.4.0-b3 has been running as a resolver for 4 weeks now after some
trouble with bind-9.4.0-b2 could have been fixed.

Since Dec 2 I can see these messages in the logs:
02-Dec-2006 12:30:21.402 general: socket.c:1173: unexpected error:
02-Dec-2006 12:30:21.415 general: internal_send: 0.0.0.1#53: Invalid
argument
...
06-Dec-2006 21:23:17.790 general: socket.c:1173: unexpected error:
06-Dec-2006 21:23:17.790 general: internal_send: 0.0.0.1#53: Invalid
argument

...
 1162         /*
 1163          * The other error types depend on whether or not the
 1164          * socket is UDP or TCP.  If it is UDP, some errors
 1165          * that we expect to be fatal under TCP are merely
 1166          * annoying, and are really soft errors.
 1167          *
 1168          * However, these soft errors are still returned as
 1169          * a status.
 1170          */
 1171         isc_sockaddr_format(&dev->address, addrbuf, sizeof(addrbuf));
 1172         isc__strerror(send_errno, strbuf, sizeof(strbuf));
 1173         UNEXPECTED_ERROR(__FILE__, __LINE__, "internal_send: %s: %s",
 1174                  addrbuf, strbuf);
...

I don't know whether it depends on each other, this afternoon we saw a
too familiar behaviour (from other hardware): CPU usage was 100% on one
CPU (normal ~15%, the machine is a Dual Opteron, threading is enabled),
machine was dropping 1.000pks/s at an avg. query rate of ~4.500q/s,
cache size ~3G.

Last time (unthreaded bind-9.3.2-P1 on Pentium with HT) this indicated a
locking issue, maybe this is similar?

Kind regards
-- 
_____________________________
[Marco Schumann
[Strato Rechenzentrum AG