Skip to content

Low consensus weight and measurements for relays with high capacity

In concrete: F7755E9EB5C17ABB4887CA7750AAADB6BA665E49

Facts and things the operator checked:

  • hardware and kernel
  • other services work fine
  • unlimited hardware/network

Tor related:

  • family: big family, afaik it shouldn't affect (though paths will be built with only 1 of the relays in the family)
  • not overloaded

Things to check:

    1. BandwidthRate/Burst are at 1GB
    1. consensus weight seems correct with respect its bw_mean and observed bandwdith:

    weight: 35MB raw measurements: 25MB observed: 15MB

    1. policies:
      • is it failing to get measured as entry? no (by logs)
      • as exit? no (by logs)
    1. use the relay as entry en tor client circuit [0]
    1. use https://gitlab.torproject.org/juga/relay_bw
    1. some other torrc configuration?
    1. test cpu with chutney

Hypothesis:

  • exit policy that restrict being measured as an exit (port 443) or entry . This is not the case

[0] https://support.torproject.org/it/relay-operators/why-is-my-relay-slow/

Edited by juga