logs archiveIRC Archive / Oftc / #tor / 2015 / July / 31 / 1
buzzkill
ralphy has a bauth problem
:)
ralphy
Hello all, I have few issues that I need help with and I'm sure someone will be able to help me. The first one is related to bandwidth limits. My relay is completely ignoring my config it seems and pushing a lot more than what I'm capable of sharing
cacahuatl
Can you paste your torrc somewhere like paste.debian.net or bpaste.net ?
ralphy
here: http://paste.debian.net/287679
My relay is pushing 25-30MB/s when I have set 6MB with 8MB burst
right now the lowest I've seen it push is 13MB/s which is way over the burst I have defined
cacahuatl
not sure what it defaults to, but by KB do you intend KBits or KBytes?
ralphy
KBytes
cacahuatl
and the measurements you're seeing are definitely in {K,M}Bytes too? not bits?
ralphy
The measures I'm looking at is arm which is Mb/sec
cacahuatl
Mb is Megabits
30Mb/s == 3.75MB/s
         

ralphy
hmmm... so 40Mbits is about 5MB/s
cacahuatl
Yeah
ralphy
duh!
cacahuatl
so just measurement confusion, looks like it's keeping it's bandwidth promises :P
ralphy
is it possible to change the configuration from KBytes to something else to make sure I keep things in a more simple way to undertsnad bans=dwidth usage?
I'll come back to that myself later. I can certainly read the config sample and its comments :P
cacahuatl
er, you can specify bits/bytes explicitly in the config
ralphy
Another issue I have which is on a separate relay... it seems that bauth is not measuring my relay properly
cacahuatl
ralphy
it has been up 9 days and atlas reports 900kb/s when this relay is capable of delivering the same 6-8MB/s as my other family members
cacahuatl
the bwauths is a known thing, there are extensive threads about it on the tor-relays mailinglist that are probably more comprehensive than I could provide for you with here.
ralphy
but isn't a bit strange that my other two relays are measured just fine? I'm kind of confused with the way each one of them behaves.
The one that I I allocated more bandwidth is the one that has been basically not used at all
cacahuatl
https://lists.torproject.org/pipermail/tor-relays/2015-May/006946.html this thread will probably have more insight into the kind of problems you might be having
ralphy
or properly measured :(
thanks cacahuatl, I'll look into that because I'm under the impression that something is not working well. I mean, I have had no downtimes, in fact these servers have more than 200 days uptime and they have been measured differently to the point that not even stable flags have been set by bauth in one case
cacahuatl
There are a few similar threads, there are acknowledged issues with bwauths, mostly covered in said mailinglist threads
ralphy
indeed, it seems that bauth is not measuring relays properly
in some cases...
w Bandwidth=20 Unmeasured=1
is what I see in concensus for my relay
no wonder there isn't more traffic going through it
cacahuatl
Yeah, so basically you've not been measured, or consensus hasn't been reached on the measurements
ralphy
"This is a known problem; the bandwidth authorities need to be worked on."
"There's no timeline set yet on when this will be fixed."
so how is the network affected by all this unmeasured thousands of relays?
if it is bad to just go by what the relay advertises (even if it isn't capable), it is worst to have 1000 relays unmeasured and not being used to their full capacity :/
isis
ralphy: are all three of your relays on the same IP/machine?
ralphy
nope
they are different IPs
and different servers
isis
okay
just checking because there's a limit of 2 relays that can be run on a single IP
i'm not sure how much the bwauth problems are hurting the network
         

ralphy
it might be much more than what we see on the naked eye
because it doesn't affect relays only but also exits
isis
the bwauth operators, or aagbsn, the developer who will be working on the bwauths, probably have some statistics
ralphy
I mean, over 1000 relays unmeasured
isis
wat
where did you see 1000+ relays?
ralphy
if they happen to be in the slow side they will not make "much" of a difference but if they are 1MB/s or higher then we can certainly think about a big impact
in concensus
unmeasured=1
isis
hmm& i see 425 currently unmeasured
but that's still a lot :/
ralphy
right, I was talking about the latest on May, sorry...
I think I missed our conversation topic
as of now there are little over 400
but it is still a lot if you ask me
although I can't claim it is "hurting" the network because I really have no way to measure that
or base my claim or anything other than speculation
I'm going to lower the advertised bandwidth to 2MB to see if somehow that helps at all
isis
mike and i did a bit of hacking on it at the end of may to get it to better deal with the Unmeasured=1 nodes: https://gitweb.torproject.org/torflow.git/log/
ralphy: maybe if you find a workaround, post to tor-relays?
ralphy
will do. from the top of my head, my first two relays that are actually measured advertised a lower bandwidth and were not included as family members.
So, I'm thinking about removing it from the family member (just in case) and lower its advertised bandwidth to see if there is better concensus
if it does make a difference I will post about it
isis
as soon as I removed it from my family members it picked up speed and was measured o_O
this is either odd or just a coincidence
between I lowered the advertised speed to just 2MB/s (that's MegaBytes)
nah, it couldn't be measured just because that was changed... Maybe it was a coincidence that it was meassured at the same time I was making these changes
isis
the family member thing is probably coincidence
ralphy
how about lowering the advertised bandwidth isis?
Could that have made a difference?
I mean, this relay has been online for 9 days with no downtime and hasn't been measured over 1MB ever
surprisingly as soon as I set the advertised speed to 2MB/s it reported 1.84 MB/s at atlas
but I know it isn't updated live so I'm basically just shooting in the dark here
isis
off the top of my head& that might make a difference? if it did, though, it would probably take longer than an hour to make a difference, since the bwauths each take about two days to finish a scan (of the entire network).
ralphy
I think that I paid attention to it and messed with the values as bauth was for the first time actually measuring it
cos' nothing else could have made a difference in about an hour that I've been pocking around with it
now I'll just give it another couple days and come back to it to see if bauth does give this relay better measurement. I mean 1.84MB/s is really far from the 8MB/s I expect to push
tamiamman
WWW.EXPOSINGCOMMUNISM.COM
WWW.VHO.ORG
jager
grah been busy
exit node is still null routed
was sent a note saying they want me to only allow 80, 443, and 53 to exit
i wrote a short note in response
http://pastebin.com/bfFVcw8e
tell me, is it likely to convince them of my point of view?
if you have advice i will take it :)
cacahuatl
You should try the tor-relays list with it, you'll probably get better formulated and informed responses, in a more verbose format.
jager
I never think of email :)
cacahuatl
there have been similar incidents in the past, and some people have chimed in to offer advice or council or speak on behalf of tor exit operators with the provider to help your case.
jager
i'll take advice where i can get it
composing a mail now
Trigraph
Hello !
« prev 1 2 3 next »