Just to confirm what others are seeing as well: I'm looking at the metrics of a guard relay running Tor 0.4.9.16 right now, and it lost its HSDir flag again. Not quite as sticky a flag as I'd have thought. -Ralph
On Thursday, 12 November 2026 03:08 Ralph Seichter via tor-relays wrote:
Not quite as sticky a flag as I'd have thought.
Quote: https://lists.torproject.org/mailman3/hyperkitty/list/tor-relays@lists.torpr... On Wednesday, 28 October 2026 19:19:44 CET David Goulet via tor-relays wrote:
There is also a big change to how directory authorities assign the HSDir flag, it is now sticky, meaning it survives a relay restart. Read: 'Sticky' in quotation marks. Emphasis on: 'survives a relay restart', the same as guard flag.
A relay still needs 96 hours of uptime to get the HSDir flag the first time but once it has the flag and stays in the consensus, it keeps it. If your relay drops out of the consensus, the uptime requirement applies again.
If an attacker takes the relay offline for 2-3 hours, the relay is out of consensus. (guard & stable flag is also gone.)
All other HSDir requirements (Stable, Fast, Running, DirCache 1) still apply as before. ^^
-- ╰_╯ Ciao Marco! Debian GNU/Linux It's free software and it gives you freedom!
* boldsuck via tor-relays:
Not quite as sticky a flag as I'd have thought.
If an attacker takes the relay offline for 2-3 hours, the relay is out of consensus. (guard & stable flag is also gone.)
That's not what happened to the guard node I mentioned. It was not taken down by the attacker, and it kept all its other flags. Only HSDir was lost. -Ralph
my HSDir seem to stick so far on my 2 guard on Tor 0.4.9.16 which you can see in the attached pictures, but they got the overloaded info in metrics. Both quite new, the older one 33 days. After the first attack on the .14 version, I increased ressources as follows: 2 Cores and 5GB RAM with minimum torrc configuration. Both on the same IP since the connection is supposed to be 10 Gbit/s. Some of the attachs lasted for hours, still all flags still assigned. I found it interesting, that the RAM usage spiked some time after the the attack started, and then never went down again once the attacks were over. [image: 11-NOV-R1.PNG] [image: 11-NOV-R2.PNG] [image: 12-NOV-R2.PNG] Good luck finding the magic potion to stop this. Am Do., 12. Nov. 2026 um 22:27 Uhr schrieb boldsuck via tor-relays < tor-relays@lists.torproject.org>:
On Thursday, 12 November 2026 03:08 Ralph Seichter via tor-relays wrote:
Not quite as sticky a flag as I'd have thought.
Quote:
https://lists.torproject.org/mailman3/hyperkitty/list/tor-relays@lists.torpr... On Wednesday, 28 October 2026 19:19:44 CET David Goulet via tor-relays wrote:
There is also a big change to how directory authorities assign the HSDir flag, it is now sticky, meaning it survives a relay restart. Read: 'Sticky' in quotation marks. Emphasis on: 'survives a relay restart', the same as guard flag.
A relay still needs 96 hours of uptime to get the HSDir flag the first time but once it has the flag and stays in the consensus, it keeps it. If your relay drops out of the consensus, the uptime requirement applies again.
If an attacker takes the relay offline for 2-3 hours, the relay is out of consensus. (guard & stable flag is also gone.)
All other HSDir requirements (Stable, Fast, Running, DirCache 1) still apply as before. ^^
-- ╰_╯ Ciao Marco!
Debian GNU/Linux
It's free software and it gives you freedom!_______________________________________________ tor-relays mailing list -- tor-relays@lists.torproject.org To unsubscribe send an email to tor-relays-leave@lists.torproject.org
On 2026-11-13 05:43, Ralph Seichter via tor-relays wrote:
* boldsuck via tor-relays:
That's not what happened to the guard node I mentioned. It was not taken down by the attacker, and it kept all its other flags. Only HSDir was lost.
All of my relays that have been attacked are done so only long enough to kick them off HSDIR, they retain GUARD and STABLE after the attack.
I run a busy Tor exit which has not regained all its flags over the last two weeks. My webtunnels have been unaffected My small relay on underpowered old laptop really suffers and was set cooking to 95C from severe Tor circuit-creation load, consistent with the HSDir attack reports. "contribution", fingerprint 97AF9CDBEEFDFDF0632845A14319BFE6B4454794, Tor 0.4.9.15, regained HSDir cools but hours later day became severely overloaded. At 2026-11-15 11:52 UTC it had 8,484 established ORPort connections from 5,904 unique IPv4 addresses across 3,483 /24s, Tor at 71% CPU and 3.7 GB RSS, host temperature 95°C, and load average 3.75. Tor reported 12,631 open circuits and continuous too slow to handle this many circuit creation requests warnings, including 13,698 suppressed events in two minutes. General overload reported 260,729 dropped Ntor requests (1.2948%, threshold 0.5%). All is consistent with the distributed HSDir-targeted circuit attack. I turned it off to cool it down. Gerry
Update Updated relay contributionx to Tor 0.4.9.17 and restarted at 12:17 UTC. HSDir remained. Within five minutes it had 5,700 established ORPort connections from 4,654 unique IPv4 addresses, with Tor using approximately 70% CPU. Temperature was 6467°C and RSS approximately 731 MB. The same high-count source addresses seen before shutdown reappeared. No overload warning had yet been emitted by the fresh process, but the distributed connection attack clearly resumed immediately. Gerry -----Original Message----- From: DocGerry via tor-relays <tor-relays@lists.torproject.org> Sent: 15 November 2026 12:14 To: 'support and questions about running Tor relays (exit, non-exit, bridge)' <tor-relays@lists.torproject.org> Cc: gerard@bulger.co.uk Subject: [tor-relays] HSDir attacks I run a busy Tor exit which has not regained all its flags over the last two weeks. My webtunnels have been unaffected My small relay on underpowered old laptop really suffers and was set cooking to 95C from severe Tor circuit-creation load, consistent with the HSDir attack reports. "contribution", fingerprint 97AF9CDBEEFDFDF0632845A14319BFE6B4454794, Tor 0.4.9.15, regained HSDir cools but hours later day became severely overloaded. At 2026-11-15 11:52 UTC it had 8,484 established ORPort connections from 5,904 unique IPv4 addresses across 3,483 /24s, Tor at 71% CPU and 3.7 GB RSS, host temperature 95°C, and load average 3.75. Tor reported 12,631 open circuits and continuous too slow to handle this many circuit creation requests warnings, including 13,698 suppressed events in two minutes. General overload reported 260,729 dropped Ntor requests (1.2948%, threshold 0.5%). All is consistent with the distributed HSDir-targeted circuit attack. I turned it off to cool it down. Gerry _______________________________________________ tor-relays mailing list -- tor-relays@lists.torproject.org To unsubscribe send an email to tor-relays-leave@lists.torproject.org
Am 15.11.26 um 13:13 schrieb DocGerry via tor-relays:
My small relay on underpowered old laptop really suffers and was set cooking to 95C from severe Tor circuit-creation load,
Is there maybe a problem with the cooling system, e.g. dust or dried out thermal compound? -- Gruß Marco Muell und Spam bitte an abfalleimer2002@stinkedores.dorfdsl.de
The Tor Silence: Who Benefits When Independent Relays Are Driven Out? The attack disproportionately affects us volunteer operators, centralising Tor capacity among large professional operators, and the lack of timely guidance is undermining trust. Gerry -----Original Message----- From: DocGerry via tor-relays <tor-relays@lists.torproject.org> Sent: 15 November 2026 12:25 To: 'support and questions about running Tor relays (exit, non-exit, bridge)' <tor-relays@lists.torproject.org> Cc: gerard@bulger.co.uk Subject: [tor-relays] Re: HSDir attacks Update Updated relay contributionx to Tor 0.4.9.17 and restarted at 12:17 UTC. HSDir remained. Within five minutes it had 5,700 established ORPort connections from 4,654 unique IPv4 addresses, with Tor using approximately 70% CPU. Temperature was 6467°C and RSS approximately 731 MB. The same high-count source addresses seen before shutdown reappeared. No overload warning had yet been emitted by the fresh process, but the distributed connection attack clearly resumed immediately. Gerry -----Original Message----- From: DocGerry via tor-relays <tor-relays@lists.torproject.org> Sent: 15 November 2026 12:14 To: 'support and questions about running Tor relays (exit, non-exit, bridge)' <tor-relays@lists.torproject.org> Cc: gerard@bulger.co.uk Subject: [tor-relays] HSDir attacks I run a busy Tor exit which has not regained all its flags over the last two weeks. My webtunnels have been unaffected My small relay on underpowered old laptop really suffers and was set cooking to 95C from severe Tor circuit-creation load, consistent with the HSDir attack reports. "contribution", fingerprint 97AF9CDBEEFDFDF0632845A14319BFE6B4454794, Tor 0.4.9.15, regained HSDir cools but hours later day became severely overloaded. At 2026-11-15 11:52 UTC it had 8,484 established ORPort connections from 5,904 unique IPv4 addresses across 3,483 /24s, Tor at 71% CPU and 3.7 GB RSS, host temperature 95°C, and load average 3.75. Tor reported 12,631 open circuits and continuous too slow to handle this many circuit creation requests warnings, including 13,698 suppressed events in two minutes. General overload reported 260,729 dropped Ntor requests (1.2948%, threshold 0.5%). All is consistent with the distributed HSDir-targeted circuit attack. I turned it off to cool it down. Gerry _______________________________________________ tor-relays mailing list -- tor-relays@lists.torproject.org To unsubscribe send an email to tor-relays-leave@lists.torproject.org _______________________________________________ tor-relays mailing list -- tor-relays@lists.torproject.org To unsubscribe send an email to tor-relays-leave@lists.torproject.org
participants (6)
-
boldsuck -
gerard@bulger.co.uk -
Marco Moock -
Ralph Seichter -
Red Oaive -
Remo Näf