Trying to understand DNSSEC and BIND versions better
Mark Andrews
marka at isc.org
Fri Jun 5 04:22:36 UTC 2009
In message <eYSdnVoGu5EsGrXXnZ2dnUVZ_vudnZ2d at posted.hiwaay2>, Chris Adams write
s:
> Since I read that the root is supposed to be signed by the end of the
> year, I am just trying to understand DNSSEC support and the various
> versions of BIND a little better here, so please don't throw too many
> rocks if I ask something stupid...
>
> I run the nameservers for an ISP. For the recursive servers, what are
> the hazzards in enabling DNSSEC (once the root is signed, so no DLV
> necessary I guess)?
Once the root is signed you will be able to validation answers
where there is a unbroken chaing of trust. DLV will still be
useful for zones were the TLD isn't yet signed or there is
another break in the chain of trust.
> I know the things that generally break with
> "regular" DNS, but I don't know that with DNSSEC (I know there have been
> DLV troubles but that's it).
Not having a clean EDNS path between the validator and
authoritative server can result in validation failures.
EDNS responses are bigger that plain DNS and may result in
fagmented responses. You need to ensure that any NAT's and
firewalls are configured to handle fragments UDP responses
up 4096 bytes with a modern BIND. Any forwarders used
should also support EDNS and preferably be performing
validation as well.
Failure to re-sign a zone will cause lookups to fail.
Failure to update DS records on DNSKEY changes will cause
lookups to fail. Failure to update DLV records on DNSKEY
changes will cause lookups to fail.
"dig +cd +dnssec <query>" is your friend. This will let
you see what is failing to validate.
"dig +cd +multi DNSKEY <zone>" will provide you with the
keyids necessary to check the signatures.
"dig +cd +multi DS <zone>" will provide you with the DS
records so you can check the linkage between parent and
child. Look at the key id field.
"dig +cd +multi DLV <zone>.<dlvroot>" will provide you with the DS
records so you can check the linkage between parent and
child. Look at the key id field.
If the zone is using NSEC3 then nsec3hash can be used to
check workout in the NSEC3 records are sane.
"date -u +%Y%m%d%H%M%S" returns the system date in a form
that is easy to comare to the dates in the RRSIG records.
A understand of how DNSSEC works is useful.
Checking if you get a DNSKEY returned, without +cd, at each zone
cut is useful for working out where to examine more closely.
dig, date and a understanding of DNSSEC is all you should
need to identify a configuration error. If the keyid match
and timestamps are good and associated NSEC/NSEC3 appear
to be sane the you will most probably have found a
implementation bug.
> Currently, my servers run BIND 9.3.4-10.P1 (as patched by Red Hat in
> RHEL; we typically stick with their security patched version, since
> that's what we pay them for). What does that mean with .ORG for
> example, where NSEC3 is used? Would we just not see NXDOMAIN responses
> as validated (and what happens to unvalidated responses)? I've put in a
> request to Red Hat to update to a version that supports NSEC3 but I
> don't know what their response will be yet.
BIND 9.3 is already at EOL.
> For our authoritative servers, we'll need to set up a system to sign the
> zones. Is it expected that ISPs will sign every zone they serve, or
> just the domains we consider "important"? What kind of problems might
> be expected here?
>
> In both cases, what kind of CPU and/or RAM overhead will large-scale use
> of DNSSEC add?
> --
> Chris Adams <cmadams at hiwaay.net>
> Systems and Network Administrator - HiWAAY Internet Services
> I don't speak for anybody but myself - that's enough trouble.
> _______________________________________________
> bind-users mailing list
> bind-users at lists.isc.org
> https://lists.isc.org/mailman/listinfo/bind-users
--
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742 INTERNET: marka at isc.org
More information about the bind-users
mailing list