Hi!
I have a hidden service on Tor 0.2.9.10 and a log file filled with:
[warn] connection_edge_process_relay_cell (at origin) failed.
The warnings appear about a minute after starting up and every 30 seconds afterwards. Service eventually degrades to only accepting a fraction of connections.
I can only get the service completely online again with a restart.
To improve uptimes I have the restart running automatically every 20 minutes. I hope there is a more elegant solution to the problem.
Maybe worth saying that this only became an issue when the service started experiencing much higher traffic volume. Despite high volume the box is idling, the tor process is at about 5% CPU and there is lots of available memory.
On Tue, Apr 04, 2017 at 09:38:10AM -0700, Joe Renege wrote:
I have a hidden service on Tor 0.2.9.10 and a log file filled with:
[warn] connection_edge_process_relay_cell (at origin) failed.
Yes, that's a known and harmless issue: https://trac.torproject.org/projects/tor/ticket/16706
The log line happens when somebody tries to connect to a port on your onion service that is not configured (i.e., your onion service listens on virtual port 80, and somebody asks for virtual port 443). This event happens increasingly often these days because of people who try to build lists of onions and scan them periodically on a pre-set list of ports.
The fix is to change the flow internally so that case doesn't count as an error that needs a warning log message. But other than the hassle of seeing a bunch of them, it's harmless, which I guess is why nobody has ever gone to fix it.
The warnings appear about a minute after starting up and every 30 seconds afterwards. Service eventually degrades to only accepting a fraction of connections.
I think this 'service degrades' thing is an unrelated, separate issue.
To improve uptimes I have the restart running automatically every 20 minutes. I hope there is a more elegant solution to the problem.
Maybe worth saying that this only became an issue when the service started experiencing much higher traffic volume. Despite high volume the box is idling, the tor process is at about 5% CPU and there is lots of available memory.
Maybe your onion service is receiving more requests (bandwidth) than it can handle? Bandwidth use would be unrelated (well, not-much-related) to CPU and memory.
In any case, it sounds like something where somebody (you) will need to collect more details of what is going wrong for you.
For perhaps somewhat related tickets, see also https://trac.torproject.org/projects/tor/ticket/15251 and https://trac.torproject.org/projects/tor/ticket/15463
And if you are running a super popular onion service you might enjoy all of the scalability things that Alec and Donncha and etc are working on, like OnionBalance.
--Roger
tor-onions@lists.torproject.org
-
Joe Renege -
Roger Dingledine