named-xfer timeout again

Mark Jeftovic markjr at privateworld.com
Thu Jan 20 07:33:21 UTC 2000


At 09:53 PM 1/19/00 -0500, Kevin Darcy wrote:
>
>Is that really "pid 0", or did you "anonymize" it? If it really says "pid
0", it
>would appear that your named is munging its internal data structures,
since pid 0
>is "impossible", and as an argument to kill() usually means "send the
signal to
>me and all processes in my process group". This also explains why named
>subsequently shuts itself gracefully, since it just kill()'ed itself.
>

That's really pid 0, not anonymized (just the domain was anonymized), so
yes, that is quite odd isn't it? My theory is that it's supposed to
kill the pid of the timed out named-xfer, but some variable or something
that should hold that pid is empty or unset and then winds up being 0,
hence the kill(0) and then bye-bye named.

>The root cause is anyone's guess. Try 1) checking/correcting your
>platform-specific compile settings and recompiling, 2) reporting the bug
to ISC,
>3) cranking up your debugging level -- maybe it'll tell you something useful.
>

The funny thing is this used to happen on another server, I moved that
nameserver to another physical server, with a completely new installation
of bind (basically, I copied over the zone files and then moved the ip
alias) and the problem followed.

Apparently this has been assigned bug id #696.

-mark


----
Mark Jeftovic, <markjr at easydns.com>
easyDNS Technologies Inc. http://www.easyDNS.com/




More information about the bind-users mailing list