[7989] in testers
odd problem with domain resolution on Athena 10
daemon@ATHENA.MIT.EDU (Alex T Prengel)
Fri Apr 3 16:35:49 2009
Message-Id: <200904032035.n33KZAKQ014792@dit.mit.edu>
To: debathena@MIT.EDU
cc: alexp@MIT.EDU, testers@MIT.EDU
Date: Fri, 03 Apr 2009 16:35:10 -0400
From: Alex T Prengel <alexp@MIT.EDU>
Hi,
I've run into two independent instances of a strange problem with host domain
resolution.
1. the latest version of Tecplot (plotting app) uses a new license
server from a different vendor than the one they used before
(FlexLM). In attempting to apply our usual access limitation to MIT IP
address ranges that both license server vendors support, I'm finding
that access is being denied to legitimate Athena 10 machines (ones
that are on campus within the allowed range) if any host name on the
127.0.0.1 or 127.0.1.1 lines (other than localhost) in /etc/hosts is
not fully qualified. i.e.
127.0.1.1 zippy zippy.mit.edu
is denied, but
127.0.1.1 zippy.mit.edu
is allowed. Note that the denial happens even if both fully qualified and
unqualified forms are present, and if both are, the order doesn't matter.
For some reason, applications that use FlexLM don't seem affected by this
issue. To see the problem, run:
add tecplot
tecplot -ver 360-09
It will either launch, or fail to with an appropriate error message.
2. I just installed a new version of the Cambridge Structural Database that
use a domain-based key (it specifies allowed IP address ranges). Exact same
problem- the user is rejected as being on an unallowed IP address if a
non-fully qualified hostname is found in /etc/hosts as described above,
but allowed otherwise.
So I'm wondering if we're doing something wrong? Is it perhaps
necessary to only have fully qualified host names on 127.0.1.1 and
127.0.0.1 lines of /etc/hosts, or is it a side-effect of something
else we're doing?
One additional data point: I see the identical issue on a machine running
straight Ubuntu Intrepid with just OpenAFS, so I don't think it's anything
particular to Athena 10 customizations.
I hope we can resolve this soon. It's a serious problem as Tecplot in
particular is a much-used application, and I suspect this will come up again.
Alex
