[tor-relays] Considering setting up an exit, need advice

[Roger Dingledine]

On Fri, Jan 31, 2014 at 01:22:47AM -0700, Jesse Victors wrote:
> Thanks again guys for the help. "usuexit" is now online, and should be
> functioning properly, but there seem to be a few mystifying issues:
> 1) TorStatus marks it as "hibernating" which it clearly isn't; it's
> online and accepting connections. I'm not sure what made TorStatus think
> it was offline.

I see quite the variety of descriptors published by usuexit. Here are
the timestamps (GMT):

published 2014-01-30 19:53:38
published 2014-01-30 20:11:27
published 2014-01-30 20:24:13 (this is the one that said hibernating)
published 2014-01-30 20:24:30
published 2014-01-30 20:30:04 (and this one too)
published 2014-01-30 20:30:35
published 2014-01-30 20:38:35
published 2014-01-30 20:39:55
published 2014-01-30 20:43:50
published 2014-01-30 20:46:34
published 2014-01-30 20:49:59
published 2014-01-30 20:52:40
published 2014-01-31 01:56:36 (and this one as well)

Have you set AccountingMax in your torrc, by chance?

Also, you sure are restarting the relay a lot. You should learn how to
use 'service reload' rather than 'service restart'. :)

> 2) For the first nine hours it has moved about 12 MB of data. I would
> think it would be more than that, especially for an exit. There's a blog
> post outlining the growth rate for guards, but how about for exits?

Here's the blog post:
https://blog.torproject.org/blog/lifecycle-of-a-new-relay
You're in 'phase one' right now.

> 3) There are mysterious warnings in the log:
> 
> Jan 30 23:07:29.000 [warn] EXTEND cell received, but not via
> RELAY_EARLY. Dropping. [4 similar message(s) suppressed in last 3600
> seconds]
> Jan 30 23:07:29.000 [warn]   (We have dropped 95.24% of all EXTEND cells
> for this reason)

You don't happen to have set "ProtocolWarnings 1" in your torrc, have you?

> I'm familiar with the meaning of the "EXTEND" cell, but I'm not sure
> what RELAY_EARLY is

See
https://gitweb.torproject.org/torspec.git/blob/HEAD:/proposals/110-avoid-infinite-circuits.txt
as motivated by
http://freehaven.net/anonbib/#congestion-longpaths

My guess is that these are unofficial Tor clients you're seeing, and
you're seeing them more because all official Tor clients are ignoring
your relay until it gets out of 'phase one' (from the blog post).

--Roger