Looks like the problem should be fixed now.
Michael J. Freedman mfreed at cs.nyu.edu
Fri Dec 2 18:07:56 EST 2005
* Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Hi,
As many of you pointed out, we experienced some downtime with CoralCDN
today as its DNS servers stopped responding to requests. I identified the
problem with its DNS servers and am in the process of pushing out a fix.
You should see service start to slowly recover in the next few minutes.
For those that are interested, the problem appears to have arisen due to
some new interaction between our DNS server and changes to the PlanetLab
platform on which we run. In the past, to determine the IP address for
which to bind our servers, we would probe the local machine to figure out
its various IP addresses, then use the first (non-loopback) address that
it returns. Unfortunately, this function was now returning a new tap
interface which was not operational / routable (10.x). I just changed it
to explicitly specify the desired IP address in the configuration files on
each node.
Sorry for the downtime. Please let me know if anybody experiences
continued problems.
Thanks,
Mike Freedman
CoralCDN Project Lead
P.S. I'd like to reiterate the suggestion that server operators that
*use* Coral please subscribe to this mailing list (either -users or at
least -announce). The set of servers using Coral is certainly much
greater than the number of mailing list subscribers. If you know of any
such operators, please forward this request. It also enables faster
feedback and identification of such problems as we encountered today.
-----
www.michaelfreedman.org www.coralcdn.org
--
I may disagree with what you have to say, but I shall defend, to the
death, your right to say it. - Voltaire
