Hi,
I've been running two nodes on the same machine for some time because I was unable to get enough throughput to blow through my bandwidth budget otherwise.
However, something seems to have changed recently, since right now I'm blasting through about double my usual bandwidth (~25 MB/s) with the two tor jobs at only ~60% CPU usage each while previously it was always CPU limited. This doesn't seem to correlate with any software updates on the machine.
Here's atlas: https://atlas.torproject.org/#details/3DD2523F1B241F01D54818F327714CDA7F5423...
ciao,
On 01/13/2014 04:37 PM, Thomas Themel wrote:
I've been running two nodes on the same machine for some time because I was unable to get enough throughput to blow through my bandwidth budget otherwise. However, something seems to have changed recently, since right now I'm blasting through about double my usual bandwidth (~25 MB/s) with the two tor jobs at only ~60% CPU usage each while previously it was always CPU limited. This doesn't seem to correlate with any software updates on the machine.
First of all, if you don't limit your relay bandwidth and hit CPU limits, this will have a negative impact for Tor users. So you should always make sure that CPU (single core!) usage stays well below 100%. An indicator of a CPU-bound problem are log entries like
"Failed to hand off onionskin. Closing. Your computer is too slow to handle this many circuit creation requests!"
Judging from Atlas, your relay might have just gotten the Stable flag. For details, see https://blog.torproject.org/blog/lifecycle-of-a-new-relay
Moritz,
How would I know that it is alright to run more than one instance of Tor? Is it possible to prevent cpu use from causing VPS businesses threatening suspension?
Robert
First of all, if you don't limit your relay bandwidth and hit CPU limits, this will have a negative impact for Tor users. So you should always make sure that CPU (single core!) usage stays well below 100%. An indicator of a CPU-bound problem are log entries like
"Failed to hand off onionskin. Closing. Your computer is too slow to handle this many circuit creation requests!"
Judging from Atlas, your relay might have just gotten the Stable flag. For details, see https://blog.torproject.org/blog/lifecycle-of-a-new-relay
On 01/13/2014 08:19 PM, I wrote:
How would I know that it is alright to run more than one instance of Tor?
VPS providers usually don't restrict how many processes you run. Some don't like if you use too many resources, as overselling is part of their business model. But how would I know? When in doubt, ask. In general, don't push it too hard. You get what you pay for, and you can't expect to get unlimited resources for the price of a donut.
Moritz,
Inarticulate I may be... but my intention is to expand Tor's contribution to people power.
The question is from an ab initio in Linux, servers and Tor guts and Tor's security integrity. I am asking from the point of view of running more than one Tor thingy on one server at one ip address. It has been said it is bad to run a Tails torrent and a relay or exit on the one VPS. Plus I don't know the actual steps to setting most things up on a server by command line.
The second part was asking if there is anyway to control the cpu or whatever load draws attention to Tor. It seems Tor would be more acceptable on more VPSs if it reliably did not cause the operators problems. After being suspended for resource abuse I asked an operator to help me meet their requirements to which they said they could only do so much about it but not prevent it.
Robert
How would I know that it is alright to run more than one instance of Tor?
VPS providers usually don't restrict how many processes you run. Some don't like if you use too many resources, as overselling is part of their business model. But how would I know? When in doubt, ask. In general, don't push it too hard. You get what you pay for, and you can't expect to get unlimited resources for the price of a donut.
-- Moritz Bartl
On 01/13/2014 11:40 PM, I wrote:
The question is from an ab initio in Linux, servers and Tor guts and Tor's security integrity. I am asking from the point of view of running more than one Tor thingy on one server at one ip address. It has been said it is bad to run a Tails torrent and a relay or exit on the one VPS. Plus I don't know the actual steps to setting most things up on a server by command line.
You can run up to 2 Tor processes per IP address. Apart from obvious resource restrictions, you can run as many Tor processes on any machine as you like.
I don't see a problem with seeding Tails and running a Tor relay on the same VPS.
The second part was asking if there is anyway to control the cpu or whatever load draws attention to Tor.
Only indirectly, by limiting its bandwidth.
Hi Moritz, Excerpts from Moritz Bartl's message of Mon Jan 13 19:30:05 +0100 2014:
First of all, if you don't limit your relay bandwidth and hit CPU limits, this will have a negative impact for Tor users. So you should always make sure that CPU (single core!) usage stays well below 100%. An indicator of a CPU-bound problem are log entries like
Ah, thanks, I wasn't aware of that. I'll have to slow things down to stay within bandwidth limits now, anyway.
"Failed to hand off onionskin. Closing. Your computer is too slow to handle this many circuit creation requests!"
None of these recently, though apparently there were some in autumn. I was probably not really overle
Judging from Atlas, your relay might have just gotten the Stable flag. For details, see https://blog.torproject.org/blog/lifecycle-of-a-new-relay
That may well have happened, the machine running was quite crash prone and recently had a rare stable spell. Still, any theories on how that would reduce my per-bandwidth CPU consumption?
ciao,
On 01/13/2014 08:48 PM, Thomas Themel wrote:
Still, any theories on how that would reduce my per-bandwidth CPU consumption?
No clue. Magic is happening inside Tor. :)
Hi Thomas,
However, something seems to have changed recently, since right now I'm blasting through about double my usual bandwidth (~25 MB/s) with the two tor jobs at only ~60% CPU usage each while previously it was always CPU limited.
Maybe your are running on a VPS and the dynamically alloted CPU frequency changed? -> /cat/proc/cpuinfo
Beste Gruesse, M.
tor-relays@lists.torproject.org
-
I -
Manuel Gebauer -
Moritz Bartl -
Thomas Themel