Re: same here 9.2.4rc2 (was Re: Bind 9.3.0 - Assertion Failure)

>
> On a debian server running 9.2.4rc2 we've had it die couple times over
> the past few days with a similar error:
>
> Jan 10 20:37:02 foo named[3192]: mem.c:653:
> INSIST(ctx->stats[size].gets > 0U) failed
>
> We're wondering if its associated with a bad memory chip.
>
> -mark

Firstly. Why are you still running a RC?

Secondly the are completely different errors.

Yours is a unbalanced number of isc_mem_put() / isc_mem_get()
at a given size. More puts than gets. In your case I would
upgrade to BIND 9.2.4 and see if the error reoccurs.

The other is that the memory pointed to by name has the wrong
magic number at the start. BIND 9.3.1 has a possible fix.
BIND 9.3.1beta1 should be out shortly.

In general errors like this should be reported to bind9-bugs@isc.org
along with a stack backtrace so we have some context to work with.
Without a stack backtrace we are working completely in the dark.

Note: 'named -u' disables writing cores as a side effect of the
setuid() call.

For FreeBSD you can stop this side effect with:
sysctl kern.sugid_coredump=1

Mark

> On Tue, 11 Jan 2005, Zain wrote:
>
> > I'm running a recursive named version 9.3.0 on a Solaris 9 sparc. The
> > named logged the following messages before existing.
> > 10-Jan-2005 10:35:33.098 name.c:1215: REQUIRE((((name) != 0) &&
> > (((const isc__magic_t *)(name))->magic == ( (('D') << 24 | ('N') << 16
> > | ('S') << 8 | ('n')))))) failed
> >
> > 10-Jan-2005 10:35:33.099 exiting (due to assertion failure)
> >
> > What could be the reason for it to happen?
> >
> > Thanks & Regards,
> > --Zain
> >
>
--
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742 INTERNET: Mark_Andrews@isc.org