Hello list. This is my 1st message here, and I'm not a pro, so please excuse me if I do something wrong. I run a relay for several years now. This is my personal notebook so I only start Tor when I power it up. I set it up for 2 things: to be a node (non-exit) and to serve me as a SOCKS proxy for accessing some sites I have here blocked. My problem is, and long time has been, the following. Every 30-40-60-120 minutes (it varies) I suddenly see CPU consumption jumps to 100% because of tor process, so I need to stop/start it to settle things down. Also, when I start Tor after a period of inactivity, say, a few hours or night, I cannot start it reliably for the first time, ever. It goes up to "Bootstrapped 95% (circuit_create): Establishing a Tor circuit" log message and... 100% CPU. After the second start, I get 100% CPU with 50/50 probability, and after the 3rd, it finally is OK, most of the time. Then, this 30-40-60-120 minutes cycle begins... It happens with this "Bootstrapped 95% ..." message or a bit before that. I experienced it with many 0.3.x, 0.4.1.x versions, now with 0.4.2.x versions also. I tried not to restart tor process when it eats my CPU for several minutes but it doesn't stop, so I have to stop/start it manually. I see nothing bad in the log at its NOTICE level. My setup now is: Windows 8 Tor is 0.4.2.6 Fingerprint BF4591C167FF01B8AAE2749CD9D04913BD6743C4 I run Tor as a Service (--service install cli option; start manually) My notebook is decently powerful and so is my network connection My torrc is basically like this: SOCKSPort [port] SOCKSPolicy accept private:*,reject *:* Log notice file C:\tor-0.4.2.6\var\notice.log DataDirectory C:\tor-0.4.2.6\var ControlPort [port] HashedControlPassword [cencored] MaxMemInQueues 1 Gb Address [my real ip] Nickname skullcactus ORPort 59001 NoListen ORPort 192.168.0.84:9001 NoAdvertise DirPort 59030 NoListen DirPort 192.168.0.84:9030 NoAdvertise RelayBandwidthRate 3 MBytes RelayBandwidthBurst 6 MBytes ExitRelay 0 GeoIPFile C:\tor-0.4.2.6\Data\Tor\geoip GeoIPv6File C:\tor-0.4.2.6\Data\Tor\geoip6 ExcludeExitNodes [some exclusions] StrictNodes 1 TrackHostExits [some hosts] TrackHostExitsExpire 600 HeartbeatPeriod 1 hours Really, I want to keep my node working but it gets on my nerves :-) . Can you please help me track the problem down?
I do run a dedicated relay (middle only, non-exit, no guard) and can at least confirm that the CPU load is very "spiky" indeed. I was wondering myself in the past but never really cared about it (since it's a dedicated device anyhow). Attached chart shows just a recent example for information. ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Friday, 17. April 2020 16:11, scullcactus <scullcactus@riseup.net> wrote:
Hello list.
This is my 1st message here, and I'm not a pro, so please excuse me if I do something wrong.
I run a relay for several years now. This is my personal notebook so I only start Tor when I power it up. I set it up for 2 things: to be a node (non-exit) and to serve me as a SOCKS proxy for accessing some sites I have here blocked. My problem is, and long time has been, the following. Every 30-40-60-120 minutes (it varies) I suddenly see CPU consumption jumps to 100% because of tor process, so I need to stop/start it to settle things down. Also, when I start Tor after a period of inactivity, say, a few hours or night, I cannot start it reliably for the first time, ever. It goes up to "Bootstrapped 95% (circuit_create): Establishing a Tor circuit" log message and... 100% CPU. After the second start, I get 100% CPU with 50/50 probability, and after the 3rd, it finally is OK, most of the time. Then, this 30-40-60-120 minutes cycle begins... It happens with this "Bootstrapped 95% ..." message or a bit before that. I experienced it with many 0.3.x, 0.4.1.x versions, now with 0.4.2.x versions also. I tried not to restart tor process when it eats my CPU for several minutes but it doesn't stop, so I have to stop/start it manually. I see nothing bad in the log at its NOTICE level.
My setup now is: Windows 8 Tor is 0.4.2.6 Fingerprint BF4591C167FF01B8AAE2749CD9D04913BD6743C4 I run Tor as a Service (--service install cli option; start manually) My notebook is decently powerful and so is my network connection
My torrc is basically like this:
SOCKSPort [port] SOCKSPolicy accept private:*,reject : Log notice file C:\tor-0.4.2.6\var\notice.log DataDirectory C:\tor-0.4.2.6\var ControlPort [port] HashedControlPassword [cencored] MaxMemInQueues 1 Gb Address [my real ip] Nickname skullcactus ORPort 59001 NoListen ORPort 192.168.0.84:9001 NoAdvertise DirPort 59030 NoListen DirPort 192.168.0.84:9030 NoAdvertise RelayBandwidthRate 3 MBytes RelayBandwidthBurst 6 MBytes ExitRelay 0 GeoIPFile C:\tor-0.4.2.6\Data\Tor\geoip GeoIPv6File C:\tor-0.4.2.6\Data\Tor\geoip6 ExcludeExitNodes [some exclusions] StrictNodes 1 TrackHostExits [some hosts] TrackHostExitsExpire 600 HeartbeatPeriod 1 hours
Really, I want to keep my node working but it gets on my nerves :-) . Can you please help me track the problem down?
tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
scullcactus wrote:
My problem is, and long time has been, the following. Every 30-40-60-120 minutes (it varies) I suddenly see CPU consumption jumps to 100% because of tor process, so I need to stop/start it to settle things down. Also, when I start Tor after a period of inactivity, say, a few hours or night, I cannot start it reliably for the first time, ever. It goes up to "Bootstrapped 95% (circuit_create): Establishing a Tor circuit" log message and... 100% CPU. After the second start, I get 100% CPU with 50/50 probability, and after the 3rd, it finally is OK, most of the time. Then, this 30-40-60-120 minutes cycle begins... It happens with this "Bootstrapped 95% ..." message or a bit before that.
I've seen the same and patched like this: --- a/lib/thread/compat_winthreads.c 2020-01-24 21:59:07 +++ b/lib/thread/compat_winthreads.c 2020-04-09 08:55:04 @@ -170,6 +170,8 @@ do { DWORD res; res = WaitForSingleObject(cond->event, ms); + + SleepEx(1, TRUE); EnterCriticalSection(&cond->lock); if (cond->n_to_wake && cond->generation != generation_at_start) { ------------- That always did the trick. I reported it here: https://trac.torproject.org/projects/tor/ticket/33411 --gv
participants (3)
-
Gisle Vanem -
petrarca＠protonmail.ch -
scullcactus