Re: host not found versus server fail

From: Greg A. Woods (no email)
Date: Thu Jan 01 2004 - 13:51:40 EST


[ On Wednesday, December 31, 2003 at 21:31:08 (-0800), Roger B.A. Klorese wrote: ]
> Subject: Re: host not found versus server fail
>
> Wietse Venema wrote:
> > He's whining about SERVFAIL errors, and I point out that there is
> > no such error result in the gethostbyaddr() manual page. He keeps
> > whining about SERVFAIL anyway, so there.
>
> Yes, but if one interface returns SERVFAIL, and another returns
> TRY_AGAIN, one possible conclusion -- clearly the one he reached -- is
> that they're synonyms.

FYI gethostbyaddr() and gethostbyname() literaly cannot possibly ever
fail with h_errno set to SERVFAIL (unless they are severely broken) --
they internally translate the low-level internal resolver error code of
"SERVFAIL" into "TRY_AGAIN". Indeed SERVFAIL is not ever a valid
value for h_errno.

In modern implementations of gethostby*() an error with h_errno set to
TRY_AGAIN indicates any number of possible temporary errors, including
SERVFAIL. One must use other tools to diagnose the exact reason for a
TRY_AGAIN error, but beware that it may not be 100% reproducable, since
of course it is an indication of a "temporary" error.

One is generally better off using dedicated testing tools to diagnose a
low-level SERVFAIL too, but if you have to go that deep and you're not a
systems programmer then you're probably approaching the problem from the
wrong direction and at the wrong level.

-- 
						Greg A. Woods
+1 416 218-0098                  VE3TCP            RoboHack <>
Planix, Inc. <>          Secrets of the Weird <>







Hosted Email Solutions

Invaluement Anti-Spam DNSBLs



Powered By FreeBSD   Powered By FreeBSD