Consensus frequently reports relay down, despite relay passing traffic
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512 Hello. One of my relays, forest19 (E21B4A2C2852298186C0EA7E2F900EF03AE38D69), frequently appears offline in the consensus despite my uptime monitors and bandwidth graphs showing uninterrupted activity. About once a day to once every other day, I get a Tor Weather notice that my relay has been offline for the last 4 hours. The relay is able to ping all authorities but Sarge, and I am not seeing abnormally high packet loss. How do I troubleshoot this? Regards, forest -----BEGIN PGP SIGNATURE----- iQIzBAEBCgAdFiEEvLrj6cuOL+I/KdxYBh18rEKN1gsFAmlN4PQACgkQBh18rEKN 1gtdxg//QENLNjbb8Ukg40Cyd5vaJDw7PSmcS7XKTh1VID6ZrCdw53arHiYBfvZM Ez26SRD+4gj6//Oz+uz2A0jcBJ8fa+RewTwm4WBthiNGLs8hhyt03CN9zA/j/sm0 9VEkxORYAnjTwJmUQYZKIzfF0GdjJirfuk/daq2q5s4Uxd1tspSSTvYI/O/myE+l 7M5qXtu7OSAJynwYGySQPudt9dGgXdb7fDCATaM7gczXpkHFkSxwIrCNHGYPPL+/ Y/xewRs5rHb0TS8DRJcvDKNBSRaTTNLUQkmiRN4it7pY3w/DWkEWduTwvufMsvK3 3AOs75oaO2cs2WVkqQag12uHTtw78uzegObZl9RPaWmThrbQ8SF/co+10zUUlUa9 UXik2zgpAyQzLLb6TNmLAkNAALQUXvO6YZnZ2z6lrkiXW5ySsgmAhgKaPX9JufKz xbXJ2kkyWRt5xP4y/I5tHj+vmjhLjahmn/gy2O3S1atdTwsd5OX0sn1MzdLfsjw2 9DPaCZpvaOCXwHoQFYM5q0K+JqDNrhsgoCMbjnIhrj/J19OCCYoaLPDd4rxrMOJ8 nsl6hzxi6804hol4A9MDnWzNpAxdxFahSbEE523TfG6OaQG4rw2GyP15Ai6NK6fO Am1cgei2NDAZBquFNB7tI2gfGvle+n7boviHdaMNmlEqMhLL7Pk= =hlTa -----END PGP SIGNATURE-----
On Fri, Dec 26, 2025 at 01:12:30AM -0000, forest-relay-contact--- via tor-relays wrote:
Hello.
One of my relays, forest19 (E21B4A2C2852298186C0EA7E2F900EF03AE38D69), frequently appears offline in the consensus despite my uptime monitors and bandwidth graphs showing uninterrupted activity. About once a day to once every other day, I get a Tor Weather notice that my relay has been offline for the last 4 hours.
The relay is able to ping all authorities but Sarge, and I am not seeing abnormally high packet loss. How do I troubleshoot this?
One really helpful tool is looking at the individual votes, to see if there is a pattern of *which* dir auths can / can't reach you. Go to the bottom of https://consensus-health.torproject.org/#relayinfo and put in forest19 and click load, and we see that at this moment, you have five Running votes (from moria1, tor26, gabelmoo, dannenberg, and maatuska), and that's a majority of nine so you are considered Running, but it sure is close to not a majority. I'm running a tcptraceroute from an IP address nearby moria1 and it works, but from moria1's outbound IP address (128.31.0.39) my tcptraceroute currently fails after 212.20.21.242 -- that is, it appears something in Russia is filtering on moria1's IP address. It looks from moria1's logs like I could reach you an hour ago (at 21:31:04 EST) but the tests after that (at 21:52:24, at 22:13:44, at 22:35:04 all EST) have failed. So I am all set to vote not Running for you in the next vote, and maybe that will be enough to lose your Running flag again. Your IPv6 address is reachable the whole time. Since this is a Russian router we are talking about, one possibility is that they are censoring the public Tor relay IP addresses, but doing it only sometimes, or on some routes (which could look like sometimes as the routes change). --Roger
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512 Hello. Roger Dingledine wrote:
Since this is a Russian router we are talking about, one possibility is that they are censoring the public Tor relay IP addresses, but doing it only sometimes, or on some routes (which could look like sometimes as the routes change).
Is the relay still useful to the network? I am seeing traffic pass through it without issue. I could ask the host to move it to one of four other locations in Russia. I chose Novosibirsk (OVB1) because it is near mainland China and I figure it would be helpful to improve geographic diversity, especially with routes to China (Novosibirsk is as far away from Europe as possible). The other locations I can migrate to are Kazan (KAZ1, looking glass 176.32.33.241), St. Petersburg (LED1, looking glass 194.87.196.227), and two DCs in Moscow (MOW1 and MOW2, LG 45.140.169.150 and 195.133.10.134, respectively). Their online ping test shows that all authorities are accessible from all of their locations, at least now. On a related note, I do have a relay in Hungary which has connectivity issues with several of my other relays (4 out of 31). It passes a fair bit of traffic regularly and has no issues with connecting to any of the authorities. Would it be appropriate for me to try to ping every relay in the consensus to try to map its connectivity? I am concerned that it is harming the network by breaking the assumption that every relay can reach every other relay. But then again, I don't know how many relays it is unable to reach. Regards, forest -----BEGIN PGP SIGNATURE----- iQIzBAEBCgAdFiEEvLrj6cuOL+I/KdxYBh18rEKN1gsFAmlOZdsACgkQBh18rEKN 1gsf8A//T0OAydbQnKIY8RI1KWs/zNQHS4EfhQzxth39O2ayByYGgeUCTa+CpYqc zCnMDJ6au1bL97VDRJ8KzUq9V5/4LMMPZ43uKCd/6T/pExPScgg9XqauMbUQDP6a e233J6/z8C/NU0x/PiKUYs8hlJfRhTQ5EfnYvBW5HPw0yrkAlw6/jgyjUh7cwtI/ 3SbzHESc0uLbMzRUOxHX8GITcwge/CrrM/fWtxP39TF8bFUoEob4qP+KfyZn6d+/ KJugU5yeyOOaCsCck8w6id/AJ4EtvAsdwEOceG2ED0m/JFxBEmO/pElZNbldmLCI Q+Y/gz78CsCmljbx/X7zHXreBlwUGO4gEqEUWX4Tb5asg0ZbvG6vMc6OoPze1c2y EysW9j5J6vGLj1IRczESsjNLG9Po+5SUZWU4a8cJsttehKobrDC8Y/ruvjPZuIRM FYh+uCnLzFTHzgNMn7h7dR0EIwT/6p5Zor3FrlaUSqm8qCdGZbfgBxmUja8sPgCR d0zGNekutu4CtpPb/MDy+yHJx1j7o0BhzJOBiZBkh65yewuyfedLbAmCWXDQlcSk O+xkMQdC7XjFgFnZ2rmsaO8nJWyvKHHYxMWUGzPsDdwBSkJArzYvcnGEV+zRl8ns 1qqAYDCYJ/WIeaHRftu7CGJQqn33MhkHaQA6jwU8VnEUMadD8gs= =e7tB -----END PGP SIGNATURE-----
On 26.12.2025 10:40 forest-relay-contact--- via tor-relays <tor-relays@lists.torproject.org> wrote:
Would it be appropriate for me to try to ping every relay in the consensus to try to map its connectivity?
Ping (ICMP echo request) is not the same as a TCP connection to the ORPort. The latter is relevant. Some operators block ICMP echo request and some countries might censor the TCP connection to the ORPort, but not ICMP. -- kind regards Marco Send spam to abfall1766742026@stinkedores.dorfdsl.de
"Is the relay still useful to the network? I am seeing traffic pass through it without issue." Yes, useful, as long as traffic passes through. Appears that is happening as the relay is reporting bandwidth, 39.67 Mbit/s. How useful or is there a more useful approach? Up to you. The issue seems to be more related to at least 8 and sometimes 9 of the 9 voting directory authorities lack of ability to keep a reliable connection via ORPort probes (roughly hourly) with your relay. Reported via "Stable MTBF", also reflected in Guard and HSDir WFU. Attached screenshot. Relay uptime is also an issue, less than a day, but that's more under relay operator direct control. Source: https://metrics.1aeo.com/relay/E21B4A2C2852298186C0EA7E2F900EF03AE38D69/ Consensus details: https://metrics.1aeo.com/relay/E21B4A2C2852298186C0EA7E2F900EF03AE38D69/#con... "The other locations I can migrate to are Kazan (KAZ1, looking glass 176.32.33.241), St. Petersburg (LED1, looking glass 194.87.196.227), and two DCs in Moscow (MOW1 and MOW2, LG 45.140.169.150 and 195.133.10.134, respectively)." Very hard to predict! One approach is get the cheapest instance possible in each location and setup a relay in each location and watch how they scale traffic. Keep the best one and cancel the other ones. Can then scale up the best one as you prefer. "I am concerned that it is harming the network by breaking the assumption that every relay can reach every other relay. But then again, I don't know how many relays it is unable to reach." Better to run a relay than to not run a relay. These are relative trade-offs. Assuming all else equal: A relay that can reach part of the network is better than no relay. A relay that can reach all of the network is better than a relay that can reach part of the network. There are many factors that influence the contributions of a relay to Tor, i.e. bandwidth, geographic location, diversity of providers, upstream, location, etc. On Friday, December 26th, 2025 at 2:41 AM, forest-relay-contact--- via tor-relays <tor-relays@lists.torproject.org> wrote:
Hello.
Roger Dingledine wrote:
Since this is a Russian router we are talking about, one possibility is that they are censoring the public Tor relay IP addresses, but doing it only sometimes, or on some routes (which could look like sometimes as the routes change).
Is the relay still useful to the network? I am seeing traffic pass through it without issue. I could ask the host to move it to one of four other locations in Russia. I chose Novosibirsk (OVB1) because it is near mainland China and I figure it would be helpful to improve geographic diversity, especially with routes to China (Novosibirsk is as far away from Europe as possible). The other locations I can migrate to are Kazan (KAZ1, looking glass 176.32.33.241), St. Petersburg (LED1, looking glass 194.87.196.227), and two DCs in Moscow (MOW1 and MOW2, LG 45.140.169.150 and 195.133.10.134, respectively). Their online ping test shows that all authorities are accessible from all of their locations, at least now.
On a related note, I do have a relay in Hungary which has connectivity issues with several of my other relays (4 out of 31). It passes a fair bit of traffic regularly and has no issues with connecting to any of the authorities. Would it be appropriate for me to try to ping every relay in the consensus to try to map its connectivity? I am concerned that it is harming the network by breaking the assumption that every relay can reach every other relay. But then again, I don't know how many relays it is unable to reach.
Regards, forest
participants (4)
-
forest-relay-contact＠cryptolab.net -
Marco Moock -
Roger Dingledine -
Tor at 1AEO