[tor-talk] hidden service not reachable: Your system clock just jumped 121 seconds forward; assuming established circuits no longer work.

[Julian Yon]

On Sun, 25 Nov 2012 23:34:30 +0100
Quan <quan at tormail.org> wrote:

> On 2012-11-25 16:22, Andreas Krey wrote:
> >> p.s.: My system clock is running correctly
> > 
> > Pretty definitely not. NB: The BIOS clock isn't the system clock
> > of the operating system (which one?); this looks like your
> > system clock gets alternatively set from two different sources
> > (and to different values that are two minutes apart); or something
> > similar.
> > 
> > How often do you get these 'just jumped' messages?
> 
> my clock is running correctly. but tor process was taking 100% of the
> cpu. clock was still running correctly during this load but tor
> somehow didn't get that.

What's your basis for “my clock is running correctly”? It's going to be
hard for anybody to diagnose your problem with so little detail:
information like what OS you're running, your hardware configuration,
other things running on the machine that could be affecting performance
are all clues to the puzzle.

One clue though is that you say Tor was at 100%. Assuming this is a
Unix-like system, what's the loadavg when this happens? (Sorry, I don't
know how to retrieve an equivalent statistic on Windows.) If it's
greater than the number of CPU cores for any extended period you have
then your problem may simply be that your computer isn't powerful
enough to do what you're asking of it.

Consider this scenario: Tor is working its arse off handling too many
circuits and/or hidden services, maxing out your CPU. Of course, Tor
isn't the only process being scheduled: there are httpd processes
serving your hidden sites and possibly non-hidden ones, other daemons
such as sshd, ftpd, smbd etc, and of course anything you're doing
interactively on the console at the time. And if your CPU can't keep
up, maybe the filesystem can't either, causing processes to get stuck
while other processes are in IOwait. Throw in memory shortage and the
associated heavy swapping (= more disk IO, more CPU usage, more
scheduling delays) and you have a recipe for things spiralling out of
control. Conceivably there could be a significant time between Tor
being scheduled out and back in again. Multiply this by a few
scheduling cycles, and perhaps it could indeed look to Tor as though
the clock had jumped forwards.

> no idea what is/was going on. I restarted tor at least dozen times,
> and everything started working normal last time I restarted tor.
> 
> problem was maybe that I had around 70 hiddenservicedir lines defined
> in torrc. could it be too much? is there a limit?

Well, everything I wrote above is purely theoretical, but if you are
running 70 hidden services on an overloaded machine that obviously adds
a little more weight to the theory. Try reducing the load on your
machine (e.g. by switching off the hidden services and other running
daemons), or running the same services on a more powerful server, and
see if it makes the problem go away.


Julian

-- 
3072D/F3A66B3A Julian Yon (2012 General Use) <pgp.2012 at jry.me>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 230 bytes
Desc: not available
URL: <http://lists.torproject.org/pipermail/tor-talk/attachments/20121126/71914295/attachment.pgp>