Opened 5 years ago

Closed 2 years ago

#13450 closed defect (worksforme)

BwAuth is leaving ~10% of relays unmeasured

Reported by: cypherpunks Owned by: aagbsn
Priority: Very High Milestone:
Component: Core Tor/Torflow Version:
Severity: Normal Keywords: performance
Cc: tim@…, teor Actual Points:
Parent ID: Points:
Reviewer: Sponsor:

Description

From Sebastian:

I started a new relay last night. As of now it is still
Unmeasured in the consensus, even tho a full scan of the network should
have occurred by now. This is concerning, because we have a large
fraction of relays (~500) which are all Unmeasured. Many of these have
the HSDir flag, indicating at least a day of uptime, some even have the
stable flag.

Damian wrote a script to parse votes and calculate the nubmer of unmeasured:

% python scrap.py
Getting gabelmoo's vote from
http://212.112.245.170:80/tor/status-vote/current/authority:
  5935 measured entries and 1332 unmeasured
Getting tor26's vote from
http://86.59.21.38:80/tor/status-vote/current/authority:
  5735 measured entries and 1690 unmeasured
Getting moria1's vote from
http://128.31.0.39:9131/tor/status-vote/current/authority:
  6647 measured entries and 625 unmeasured
Getting maatuska's vote from
http://171.25.193.9:443/tor/status-vote/current/authority:
  6313 measured entries and 1112 unmeasured

Child Tickets

Attachments (2)

2015-01-01__2015-05-13_unmeasured_count.csv (55.7 KB) - added by cypherpunks 5 years ago.
This csv file shows you how many relays had 'Unmeasured=1' per consensus
unmeasured_relaycount.png (116.7 KB) - added by cypherpunks 5 years ago.
this is the graph made out of the csv file (2015-01-01 - 2015-05-13)

Download all attachments as: .zip

Change History (10)

comment:1 Changed 5 years ago by cypherpunks

cat bwscan.V3BandwidthsFile | grep -v "bw=1 "|wc -l
8621

cat bwscan.V3BandwidthsFile | grep "bw=1 " | wc -l
121

121/8741
.0138

Can you post these #s?

comment:2 Changed 5 years ago by tsammut

Cc: tim@… added

comment:3 Changed 5 years ago by teor

Cc: teor added

comment:4 Changed 5 years ago by teor

Do we know whether it's the same relays which are unmeasured each time?

Reports on tor-relays seem to suggest that particular relays remain unmeasured for long periods of time. Some relay operators have found that changing certain characteristics (key? exit flag?) can get their relay measured again.

comment:5 Changed 5 years ago by tyseom

Cc: tyseom added

Changed 5 years ago by cypherpunks

This csv file shows you how many relays had 'Unmeasured=1' per consensus

Changed 5 years ago by cypherpunks

Attachment: unmeasured_relaycount.png added

this is the graph made out of the csv file (2015-01-01 - 2015-05-13)

comment:6 Changed 5 years ago by cypherpunks

note also the decrease in advertised bw since the last two weeks:
https://metrics.torproject.org/bandwidth.html

comment:7 Changed 3 years ago by tyseom

Cc: tyseom removed
Severity: Normal

comment:8 Changed 2 years ago by teor

Resolution: worksforme
Status: newclosed

This is not a live issue any more.

Note: See TracTickets for help on using tickets.