Re: tor server thinks IP address changed twice when it did not change

[Scott Bennett <bennett@xxxxxxxxxx>]

     On Wed, 21 Nov 2007 20:52:14 -0500 Roger Dingledine <arma@xxxxxxx> wrote:
>On Wed, Nov 21, 2007 at 09:11:36AM -0600, Scott Bennett wrote:
>>      However, a while later and completely unnecessarily, tor did this:
>> 
>> Nov 21 07:56:52.311 [notice] Your IP address seems to have changed. Updating.
>> Nov 21 07:56:53.403 [notice] Self-testing indicates your ORPort is reachable from the outside. Excellent. Publishing server descriptor.
>> Nov 21 07:57:55.025 [notice] Your IP address seems to have changed. Updating.
>> Nov 21 07:57:59.599 [notice] Self-testing indicates your ORPort is reachable from the outside. Excellent. Publishing server descriptor.
>> Nov 21 07:58:02.279 [notice] Self-testing indicates your DirPort is reachable from the outside. Excellent.
>
>Interesting. Looks like it changed, then changed back. I just made

     Thanks for writing back.
     That's what it looks like, except that a) it didn't change, and b) even
if it had, it would take up to ten minutes for inadyn to check it, find it had
changed, and then update the name servers, and then tor would take several more
minutes to notice that the name resolved to a new address.  ~61.5 seconds is
just not long enough.

>the "seems to have changed" log entry more informative:
>http://archives.seul.org/or/cvs/Nov-2007/msg00246.html
>
>> Nov 21 07:58:16.929 [notice] Have tried resolving or connecting to address '[scrubbed]' at 3 different places. Giving up.
>> Nov 21 07:58:16.965 [warn] We just marked ourself as down. Are your external addresses reachable?
>
>And these two are from the dirport reachability test launched at
>07:56:52.
>
>>      Meanwhile, the connection was good the whole time, tor traffic was being
>> relayed, and the IP address remained unchanged since the previously noted
>> event.  In the slightly more than an hour that has passed since the weirdness
>> occurred between 07:56 and 07:59, tor has continued to run, answering directory
>> requests and passing traffic.
>
>Right. That's because Tor clients were using the IP address advertised
>in the descriptor you published before this hiccup.

     The thing is, if the address really had changed, all the open connections
would be dead meat, just as happens every time my f***ed up ISP forces a
disconnection and then assigns a new address when the router reconnects. :-(
>
>>      So it looks like there is a bug involved here, and it apparently does
>> not depend upon starting up with no cached-descriptors* files.
>
>Right, there's no reason to think it has to do with your cached files.
>Tor doesn't look at your old cached descriptors to decide your current
>IP address -- that would just be too fragile.

     The only reason I thought it might be related was that 0.2.0.12-alpha
began screwing up right away when it was started without those files, and
at the time at least, it didn't screw up right away when started with those
files.  Apparently, that was just a matter of coincidental circumstance.
>
>>  Does anyone
>> working on tor development recognize these symptoms?  Or is this something new?
>
>My first thought is that in fact your "thruhere.net" address resolved
>to the old IP address briefly. This might happen because thruhere.net
>has a big pile of dns servers, if one of them had an old answer cached

     Hmmm...yes, it looks like there are five NS RRs for it.

>for a little while.

     Well, the only problem with that is that the TTL on the A RR is only
60 seconds, so the wrong address should not have been returned more than
an hour after inadyn had updated the address on the name servers.
>
>How repeatable is this? If you could get an info-level log, that would

     I have no idea.  It did it two runs in a row, but once was right away,
and once was nearly a day after startup.

>give some more hints. (A debug-level log would provide even more hints,
>but it would be really loud, so let's first see if info is enough.)
>
     Okay.  I have it logging at "info" for now.  I hope it does it soon,
though, because even info-level logging is quite voluminous, as I discovered
on 0.2.0.9-alpha's "out of date" problem.


                                  Scott Bennett, Comm. ASMELG, CFIAG
**********************************************************************
* Internet:       bennett at cs.niu.edu                              *
*--------------------------------------------------------------------*
* "A well regulated and disciplined militia, is at all times a good  *
* objection to the introduction of that bane of all free governments *
* -- a standing army."                                               *
*    -- Gov. John Hancock, New York Journal, 28 January 1790         *
**********************************************************************