[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: [tor-talk] [tor-consensus-health] Consensus issues



17.08.2014, 00:27 Sebastian Hahn:

@Damian Johnson (to avoid sending two emails)

I wouldn't have worried either if I wouldn't have found it strange to
see the notice for tor26. Please read on below.

> thanks for looking after the network!

The doctor is doing that for me. I just wanted to figure out if its
diagnose is correct.

Thank you for operating gabelmoo, thank you for your support, thank you
for...

> On 16 Aug 2014, at 22:56, Sebastian G. <bastik.tor> wrote:
>> On Sat, 16 Aug 2014 19:46:15 +0000 (UTC) the doctor said:
>>> NOTICE: Consensus belonging to maatuska was missing the following authority signatures: tor26
>>> NOTICE: Consensus belonging to tor26 was missing the following authority signatures: tor26
>>> NOTICE: Consensus belonging to urras was missing the following authority signatures: tor26
>>> NOTICE: Consensus belonging to dizum was missing the following authority signatures: tor26
>>> NOTICE: Consensus belonging to gabelmoo was missing the following authority signatures: tor26
>>> NOTICE: Consensus belonging to moria1 was missing the following authority signatures: tor26
>>> NOTICE: Consensus belonging to dannenberg was missing the following authority signatures: tor26
>>> NOTICE: Consensus belonging to Faravahar was missing the following authority signatures: tor26
>>
>> If I understand this messages correctly tor26 didn't sign the consensus
>> of any other authority. (Correct me if I'm wrong.)
>>
>> How is it possible that tor26 doesn't sign its own consensus?
> 
> Here's an easy theory on what might have happened: When it was time to
> vote, tor26 made a vote, and distributed it to the other dirauths. When
> it was done doing so, it went offline. The other dirauths made a
> consensus, and signed it. tor26 came back online, learned that there was
> a consensus it didn't know about, fetched it from the other dirauths,
> but didn't sign it - because the time to sign it was in the past. This
> does not constitute an error condition for tor26, because enough other
> dirauths signed it for it to be considered valid.

That appears to be logical. This is also in line with what I believed to
know about how the consensus is formed.

It was indeed the wording of the notice line(s) that made me believe a
consensus of an authority would somehow be bound to that authority.

> I'd argue against increasing the complexity of the voting process to
> handle this rare edge case. I do think maybe the wording is confusing:
> What does "Consensus belonging to" mean? A consensus doesn't belong to
> any individual dirauth. I don't have a quick suggestion for what to
> name the notice instead, tho.

Attention Damian Johnson ;)

How about "Consensus fetched from ..." or "Consensus downloaded from..."?

(If that requires discussion I happily open a ticket)

I don't see any benefit (at first glance) to change (and most likely
increase complexity) the voting/signing process. Especially since the
consensus is valid and it is expected to be fetchable from any authority.

>> A similar message was send on the 15th for gabelmoo, but gabelmoo had no
>> notice line. There were two warning, first gabelmoo did not publish a
>> fresh consensus and secondly it did not report bandwidth scanner
>> results. Nothing I would have worried about. Nor would I have found strange.
> 
> Yes, gabelmoo was down as I was fixing its bw auth. Nothing to worry
> about indeed.
> 
>> However an authority handing out a consensus it didn't sign might be
>> something that isn't quite right.
> 
> I think it's OK, considering the above.

When it is what it appears to be then I agree.

Thank you both for your replies-

> Cheers
> Sebastian
Sebastian G.


-- 
tor-talk mailing list - tor-talk@xxxxxxxxxxxxxxxxxxxx
To unsubscribe or change other settings go to
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk