Re: [tor-relays] Questions about consensus votes

2021-04-22 Thread Tobias Höller

On 21.04.21 18:17, Toralf Förster wrote:

On 4/21/21 12:15 PM, Sebastian Hahn wrote:
Moria applies its own criteria which differ from dir-spec. Its 
operator is testing future improvements to the Tor network and 
therefore frequently doesn't follow all the specs.

bastet too, or ?


Yes, bastet and longclaw do also not vote according to dir-spec. To be 
fully transparent, longclaw used to not vote according to dir-spec (at 
least for March 2021), they changed their voting behavior on March March 
29 2021 13:00, since then they have started voting according to dir-spec 
again. Which kind of brings me to my next question: Are all dirauths 
allowed/expected to ignore the dir-spec and test other voting 
configurations?


Tobias

___
tor-relays mailing list
tor-relays@lists.torproject.org
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays


Re: [tor-relays] Questions about consensus votes

2021-04-21 Thread Toralf Förster

On 4/21/21 12:15 PM, Sebastian Hahn wrote:

Moria applies its own criteria which differ from dir-spec. Its operator is 
testing future improvements to the Tor network and therefore frequently doesn't 
follow all the specs.

bastet too, or ?

--
Toralf
___
tor-relays mailing list
tor-relays@lists.torproject.org
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays


Re: [tor-relays] Questions about consensus votes

2021-04-21 Thread Sebastian Hahn
Hi Tobias,

thanks for running a relay!

> On 20. Apr 2021, at 15:21, Tobias Höller  wrote:
> 
> I have only recently started operating relays 
> (Family:008196DC449482C73CFA9712445223917F760921) and have some trouble with 
> reliably getting the "Fast" and "HSDir" flags for my relays. Right now they 
> regularly obtain and loose these flags and I would like them to be more 
> stable To figure out what was going wrong, I took a look at the archived 
> votes at https://collector.torproject.org/archive/relay-descriptors/votes/
> Unfortunately, I still don't understand why some things are happening, so I'm 
> hoping you can help me out:
> 
> 1. The "Fast" flag is assigned by either measurements (if the DirAuth is also 
> a Bandwidth Auth) or the reported bandwidth. On my relays I noticed that if I 
> set the "BandwidthRate" close to the fast threshold (105KB/s) I would not get 
> the fast flag, unless I configure a much higher bandwidth burst rate 
> (>5MB/s). Even more confusing, different bandwidth auths seem to require 
> different burst rates. While for example Moria1 started voting me fast with a 
> Bandwidth Burst rate of 5Mbit/s, maatuska did not do so until I increased the 
> BandwidthBurst to 100Mbit/s (my complete internet uplink). What I'd like to 
> know: Is that something that only affects me or have you also experienced it? 
> And are there any recommendations how high the bandwidth burst should at 
> least be compared to the desired bandwidth rate?

It is quite natural that different dirauths will reach different verdicts, 
because they are all located at different locations around the world. 
Additionally, the exact cutoffs for the flags are somewhat dynamic depending on 
what other relays are doing - in general, there's no hard number that will 
definitely be enough or definitely won't be.

> 2. I am confused about the HSDir flag: According to 
> https://gitweb.torproject.org/torspec.git/tree/dir-spec.txt#n2500 it should 
> require the relay to be fast+stable+ uptime>MinUptimeHidServDirectoryV2 (by 
> default 96 hours). So, if an authority votes my relay to be 
> Running+Stable+Fast for 100 consecutive votes, it should vote for "HSDir" in 
> the last 4 votes, right?. Because moria1 for example, votes my relays to be 
> fast+stable+running constantly, but never votes for the HSDir flag.
> Is there another condition for becoming HSDir that I am not aware of, or can 
> you think of an other reason why this is happening?

Moria applies its own criteria which differ from dir-spec. Its operator is 
testing future improvements to the Tor network and therefore frequently doesn't 
follow all the specs.

> Lastly, I'd be curious to learn if I am the only operator who has trouble to 
> maintain flags on their relays, or if you have also experienced similar 
> problems to mine. If yes, how did you solve them.

Yes, some relays have this problem. In general I would recommend not worrying 
about the flags so much. Provide the most useful relay you can from a 
bandwidthrate perspective, keep it online, and that's already helping the 
network a lot.

That said, it is useful to see if the network as a whole is actually acting as 
designed. You seem to have invested quite a lot of effort already into dealing 
with flags, so perhaps that would be something that interests you more. Tor 
metrics might be a starting point if you wanted to go that route

Cheers
Sebastian
___
tor-relays mailing list
tor-relays@lists.torproject.org
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays