logs archiveIRC Archive / Oftc / #tor / 2009 / December / 30 / 1
fabi
arma: alive?
(or anyone else that can help me with the bw authority scanners)?
arma
what's up?
mikeperry
fabi: I am here
fabi
ah hi!
i tried to set up the new bw authority mechanism in my private testing tor network.
My network used to have 3 authoritative dirs, but currently only two are running. I installed the bwauthority scanner on both and they seem to work fine. Both produce a well looking bwscan.V3BandwidthsFile.
But I don't see the new bw values in the consensus. There only seem to be the old bw-observed values from the descriptor..
i have no clue what's going wrong :(
babadadada
phobos: did you know there is a way to proxy Java, ActiveX etc.., to make sure that everything goes through your proxy of choice?
nsa
or: arma committed revision 21356 (/projects/misc-sysadmin): found these lines in my version of captain's log
phobos
babadadada: iptables tricks
basically redirect all tcp traffic through tor
https://wiki.torproject.org/noreply/TheOnionRouter/TransparentProxy
babadadada
babadadada: I use the internet unrestrictedly with my proxy
lol
phobos: : I use the internet unrestrictedly with my proxy
phobos: do you do so with the "TransparentProxy" as well?
phobos: this will redirect ALL traffic through tor?
phobos
i think so
         

babadadada
"i think so"
dr|z3d
All TCP traffic.
Other protocols will be dropped.
babadadada
dropped?
that doesnt sound good
dr|z3d
"go nowhere fast"
misc
ah ah
fabi
mikeperry / arma: no idea? too less information?
dr|z3d
Tor only supports TCP. UDP doesn't work.
babadadada
dr|z3d: will UDP /ever/ work?
dr|z3d
babadadada: Only when support has been implemented.. don't hold your breath!
babadadada
thats a bummer
dr|z3d
If you need UDP support, perhaps have a look at I2P as your companion-to-Tor app.
babadadada
i thought tor was just one of those net-within-the-nets
i mean I2P
dr|z3d
It is, mostly. A perfect compliment to Tor.
babadadada
well what if i want to make udp connections to the normal internet?
dr|z3d
You'd need a UDP "outproxy".
babadadada
where would I find that?
dr|z3d
It's theoretically possible; whether there's a "public" udp outproxy already set up I cannot tell you.
babadadada
where does I2P come into play here?
dr|z3d
It supports protocols Tor doesn't, that's where.
Anyways, it was a pointer, not an invitation to go into the finer details of I2P. Ymmv. Further I2P chat -> #nottor. :)
babadadada
yes but only for its own "inner internet," unless I'm mistaken, right?
         

dr|z3d
...
babadadada
I've joined. Please respond to my question in #nottor
fabi
hrm... :(
arma
fabi: reading
fabi: your authorities should be writing out a v3-status-votes file
that should give you some hints.
fabi
where do i find it?
btw. are 2 authorities enough?
ah, i found the file
both files don't give me a hint, that there is something going wrong...
arma
does the directory authority torrc file specify to use the bandwidths file?
fabi
yes
arma
do its logs indicate that it's reading it?
fabi
no
on which log-level should i see it?
arma
reenable that log line and see about then? :)
fabi
what log line?
arma
reading code is required for this feature, if it doesn't work out of the box
dirserv_read_measured_bandwidths()
fabi
how could i find the problem if there's even nothing logged at all?
arma
your v3-status-votes file. does it have any measured= entries to it?
fabi
yes
arma
what tor version are the authorities running
fabi
0.2.2.5
-alpha
w Bandwidth=136 Measured=67
arma
you need a majority of measured lines before it'll switch to the median measured value
fabi
what is a "majority" of 3 authorities with only two of them running?
arma
2
(majority of 3 is 2)
fabi
that's what i thought
so if 2 authorities vote a measured value for more than 50% of all OR, there must be at least one in the consensus using the measured value, right?
arma
it's all n separate decisions, for n relays
note that the consensus doesn't contain any Measured= strings. it just has the correct number.
fabi
yes, but there should a difference between the value of consensus and descriptor
arma
yep
fabi
and i don't see a difference on any router
can you tell me, what this means:
WARN [ Wed Dec 30 00:03:05 2009 ]: Bandwidth scan stale. Possible dead bwauthority.py. Timestamp: Tue Dec 22 23:29:38 2009
bwauthority.py is not dead, i restarted it several times
and the authority produces correct v3-status-votes file
arma
that is your clue
tor doesn't want it. its timestamp is too old.
fabi
that was my first thought
arma
(well, maybe. the correctly formed votes may mean it's ok)
fabi
how can the timestamp expire that way? The scanner's still running..
(and has been restarted repeatedly)
btw, this log is produced by scanner's cron.sh, not tor
hmm. sorry, i fall asleep on my keyboard ;) maybe we can continue tomorrow, thank you anyway!
nsa
or: nickm committed revision 21357 (/bridgedb/trunk): Remove everything in bridgedb but README_BRIDGEDB_HAS_MOVED
or: nickm committed revision 21358 (/bridgedb): Remove everything in bridgedb but README_BRIDGEDB_HAS_MOVED
or: nickm committed revision 21359 (/torbutton/trunk/trans_tools): Remove the pootle external from svn.
or: nickm committed revision 21360 (/torbutton/trunk): Remove torbutton trunk with a README_torbutton_has_moved
or: nickm committed revision 21361 (/torbutton/branches/torbutton-1_2_x/trans_tools): Remove the pootle external from torbutton 1.2.x svn branch
or: nickm committed revision 21362 (/torbutton/branches/torbutton-1_2_x): Remove torbutton 1.2.x branch with a README_torbutton_has_moved
or: arma@seul.org committed patch by Roger Dingledine <arma@torproject.org> at Tue, 29 Dec 2009 22:26:01 -0500 (EST) to tor/master: add separate per-conn write limiting
arma
wonder what that'll break
nsa
or: arma@seul.org committed patch by Roger Dingledine <arma@torproject.org> at Tue, 29 Dec 2009 23:13:26 -0500 (EST) to tor/master: add config options to override.
miraiw
bye
iledoirchirc
hi
what time in virginia?
in bombay?
is it possible to reveal the connection to TOR, mybe with vpn?
no experts here?
nsa
or: mikeperry committed revision 21363 (/torflow/trunk/CircuitAnalysis/BuildTimes):
or: Initial commit of CBT stress testing tools.
Sebastian
gouki: ping?
gouki
Sebastian, here mate!
Sebastian
gouki: hey, sorry, I've not been productive wrt the manpage conversion so far. Are you working on that beast? Else I'll do some stuff.
gouki
Sebastian, sorry. I've been WASTING time playing with my Android phone :)
Sebastian
well, if you want to play with it, I'll convert the tor-gencert manpage.
gouki
Sebastian, that would be cool.
Sebastian
gouki: ok, looking forward to your next commit.
gouki
Sebastian, sure thing buddy. I'll be able to have it pretty close to concluded today or tomorrow morning.
Sebastian
right, then we'll have to fix the Makefile and other fun stuff. And then make it make more sense. Your help so far has been most wonderful
gouki
No problem. Doing what I can to help.
(Action) stops playing with cowsay on Android
Sebastian
:)
gouki: did you figure out how to do the underline:underline stuff?
gouki
Sebastian, no. Have you? I was told that it wasn't possible because AsciiDoc does not have text decoration ... o.0
Sebastian
gouki: no, I haven't. But I'm sure it is possible.
hrm
(Action) tries to learn how to do that first.
fabi
Hello again..
SwissTorExit
hi fabi
fabi
Coming back to yesterday's discussion.. I still have the problem with setting up the bwauthorities on my private testing network
Both running auth-dirs have scanners running and seem to produce valid files - but they don't include the new values in the consensus.
My only clue is: On one of them I get the following log, when i run cron.sh:
WARN [ Wed Dec 30 13:16:59 2009 ]: Bandwidth scan stale. Possible dead bwauthority.py. Timestamp: Tue Dec 22 23:29:38 2009
SwissTorExit
well for such think , i can't help you but someone for sure ;)
fabi
I can't imagine a reason. The scanner is running and has been restarted several times
Sebastian
fabi: maybe it cannot write to the file, or something? dec 22 was a while ago.
fabi
SwissTorExit: Gr├╝eziwol! Yes, I think the number of people being able to help is quite limited :(
SwissTorExit
hi Sebastian :P
fabi
Sebastian: I think this can't be since nothing's been changed since 22.12.
SwissTorExit
but sure happy to help for sure thing eheheh
fabi
if it cannot write the file now it shoud never have written it
Sebastian
well, that date is coming from somewhere, and if everything worked as it should work, there wouldn't be a problem ;)
fabi
exactly
so why doesn't it work again after I restart all scanners and the tor client?
on my other authority this isn't a problem
Sebastian
well, what did you find when you analyzed cron.sh to figure out where that date comes from?
fabi
both are completely identical
i think it came from aggregate.py?
Sebastian
I don't know, but I'm sure grep can tell you
what I'm trying to tell you is that you're the one running this thing with the ability to debug
you should make sure you do that as much as you possibly can
fabi
that's not as easy as it sounds, since it's mostly undocumented
Sebastian
right
well, hrm
I guess I shut up, and your problem remains unsolved
fabi
scan_age = int(round(min(scanner_timestamps),0))

if scan_age < time.time() - MAX_SCAN_AGE:
,
that causes this log
I mean - I would understand, if the scanner results _are_ that old... but they aren't
Sebastian
right, so you've found a bug
see, you really really really need to understand that when there is a bug, stuff doesn't work like it *should* work. So you have to challenge your assumptions
maybe the filesystem reports the wrong date? Have you checked? Maybe the rounding code doesn't work, have you added additional logging?
etc
You're in the unfortunate situation that you have some time pressure, and the author and the only users of that code have no time/interest to help you debug it currently, so you have to do it yourself
by analyzing the problem and giving concrete pointers, you might help them help you
fabi
it is very very unlikely to have any file system or date reasons
Sebastian
Right
fabi
for simplicity imagine both authorities are working on a identical VM. One works, the other not
"some" time pressure is funny ;)
Sebastian
That by definition means that they are not identical
fabi
meine fresse
Sebastian
seriously
what do you expect from us?
fabi
they are not running on the same machine, that is the only difference
Sebastian
well, they might have a random number generator, calling "should_i_work()"
fabi
clocks are syncronized
Sebastian
seriously, what do you expect?
fabi
any hint, where to find a solution, that is a little more detailed than "look at the code"
Sebastian
If any of us knew the solution off the top of our heads, we would have provided it. I promise.
dr|z3d
fabi: You might need to wait for other devs to appear who have a more intimate knowledge of the code in question. nickm for example.
Sebastian
dr|z3d: except nickm didn't write any of that code
fabi: I personally don't run the scanner.
I just looked at the code for the first time
dr|z3d
Sebastian: He may have more intimate knowledge, however. Maybe not. The point is, others _might_ have a better view of things, but are currently absent. :)
fabi
yes. I think we were very close to a solution yesterday..
Sebastian
fabi: For example, why not add some debug code to line 233, to see if it opens files
fabi: and print the timestamp right there
also it looks like the code emits a warning when it doesn't need to do so
So maybe the problem lies elsewhere
« prev next »