Re: [Tails-dev] [review'n'merge:1.2] feature/tor-0.2.5.x-alp…

Supprimer ce message

Répondre à ce message
Auteur: intrigeri
Date:  
À: The Tails public development discussion list
Sujet: Re: [Tails-dev] [review'n'merge:1.2] feature/tor-0.2.5.x-alpha (#7907) [Was: Tor 0.2.5.x and Tails]
hi,

anonym wrote (25 Sep 2014 17:26:57 GMT) :
> Now I get the exact same issue as I reported yesterday privately: about
> half of the times tor 0.2.5.8-rc gets stuck during bootstrap for a long
> time (perhaps indefinitely sometimes). There also seem to be
> post-bootstrap issues that can result in circuits timing out. When
> testing the image in the automated test suite, the delay often exceeds
> the maximum time we allow in the "Tor is ready" step, resulting in test
> failure, but sometimes it "just" adds about two minutes of extra run
> time for the feature/scenario.


> [...]


> Because of [1] this also affects (well, affected [2]) the devel branch
> so the issue has nothing to do with enabling the sandbox. In fact, I've
> also tried both tor 0.2.5.7-rc and 0.2.5.8-rc in a 64-bit Debian Sid VM,
> and as long as I clear /var/lib/tor to force a full bootstrap, I can
> reproduce this problem, although it seems a lot rarer, like 10%.


> Since I've been running 0.2.5.x for over a month without issue, I wonder
> if the problems (both during and after bootstrap) only arises when a
> full bootstrap is done. Well, I don't know what to make of all this, so
> can someone please try reproducing this?


I've built an ISO from experimental (Tor 0.2.5.7-rc), booted it 10
times with a pretty fast and stable Internet connection, and:

* 9 times, Tor bootstrapped reasonably quickly
* 1 time, it took 5 minutes to go from 50% to 100%

This doesn't confirm the "half of the times" failure rate, but it's
indeed bad enough to be worrying. I'll try to find time to do the same
test tomorrow, on the same Internet connection, with Tails 1.1.2 and
in the automated test suite.

It seems that everytime there's a problem, connections to
[scrubbed]:443 are attempted during bootstrap. I'm assuming that these
are client connections. Might it be that 0.2.5.x is badly affected by
circuit requests while bootstrapping?

IMO, next step is to report a bug upstream, so that Nick, Andrea or
Roger can have a look at it.

Cheers,
--
intrigeri