Re: Re: Random name-resolution failures since glibc-2.1.94 upgrade?

[<tnvander@chello.nl>]

> 
> Van: James Antill <james@and.org>
> Datum: 31 Oct 2000 12:49:46 -0500
> Aan: <tnvander@chello.nl>
> Onderwerp: Re: Random name-resolution failures since glibc-2.1.94 upgrade?
> 
> <tnvander@chello.nl> writes:
> 
> > Hello,
> > 
> > Since the introduction of glibc-2.1.94 into Woody I randomly get name-resolution failures. Not all applications seem to be affected equally, it seems that apt-get is having problems more often than other software but it's not limited to apt-get.
> > 
> > The problem appears to be completely random, and comes and goes. It occurs with both 2.2.x and 2.4.xtest kernels.
> > 
> > Name resolution never really fails for all sites; I might be perfectly able to resolve www.debian.org for instance, while totally unable to resolve ftp.uk.debian.org. And then after a while, perhaps ten minutes, perhaps more, sometimes less, it works again.
> > 
> > I have never seen this kind of problem until the glibc upgrade; name-resolution sometimes failes due to DNS weirdness of my provider but never in this way.
> > 
> > 
> > In case it's of any help, I attached ltrace and strace output for telnet.
> > 
> > 
> > Does anyone experience similar problems? Or have any clues as to what I can do to fix the problem or debug it?
> 
> [snip ... ]
> 
> > open("/etc/resolv.conf", O_RDONLY)      = 3
> > fstat(3, {st_mode=S_IFREG|0644, st_size=33, ...}) = 0
> > old_mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x40017000
> > read(3, "search bogus\nnameserver 10.0.0.1"..., 4096) = 33
> > read(3, "", 4096)                       = 0
> > close(3)                                = 0
> 
>  This is a local nameserver then (Ie. you are running it on the same
> machine ?). I'm guessing yes, so I'd look for the problem here first.


This is indeed a local nameserver, and it does sometimes seem to fail. However, it's failure is very sporadic and even if it does fail, I'm wondering if it's really the failure of the local nameserver or the failure or my ISP's nameservers.

I don't think that my problems are related to this local nameserver, as it's running on another machine which is not a Debian box (it works, and I never get round to converting it;). No configuration changes or software updates have been made to that box for a long time.
That's one of my reasons for not seeking the reason there. Other reasons are that nameresolution-failures caused by the DNS server failing usually give different error messages, and that these specific error messages started occurring right after the upgrade to glibc-2.94 (and haven't yet disappeared with 2.95).

To make sure, I can test using another box when this happens again.


> Has bind been recompiled since the glibc upgrade as I know it was
> said that all applications that were compiled with ipv6 support have
> to be recompiled (the ipv6 committe changed the struct for ipv6), so
> this _might_ be a problem.
>  If not then an strace of bind as you telnet might provide some
> answers.
> 
> [snip ... ]
> 
> -- 
> James Antill -- james@and.org
> "If we can't keep this sort of thing out of the kernel, we might as well
> pack it up and go run Solaris." -- Larry McVoy.
> 


Thanks for the answer,

--Tim