Tor Metrics 'Running' flag is back for bridges who don't publish the OrPort

I don't know if this was mentioned at the Tor Relay Meeting yesterday, if so, I missed it. ;-) A few months ago there was a recommendation to not exposing OrPort for bridges. This had the unpleasant effect that all bridges were 'red' on Tor Metrics, even though they were running perfectly fine. I noticed yesterday after the meeting that everything is 'green' again. https://metrics.torproject.org/rs.html#search/ForPrivacyNET Thank you, I believe these 6 people did that: https://gitlab.torproject.org/tpo/network-health/team/-/issues/318 -- ╰_╯ Ciao Marco! Debian GNU/Linux It's free software and it gives you freedom!

Oh, the fix only lasted 23 hours. ;-) 'Running' flag is gone again. -- ╰_╯ Ciao Marco! Debian GNU/Linux It's free software and it gives you freedom!

On Sun, Jun 23, 2024 at 09:28:23PM +0200, lists@for-privacy.net wrote:
A few months ago there was a recommendation to not exposing OrPort for bridges. This had the unpleasant effect that all bridges were 'red' on Tor Metrics, even though they were running perfectly fine. I noticed yesterday after the meeting that everything is 'green' again. https://metrics.torproject.org/rs.html#search/ForPrivacyNET
Thank you, I believe these 6 people did that: https://gitlab.torproject.org/tpo/network-health/team/-/issues/318
Right -- I think we are still in the process of fixing that issue. The current situation as I understand it is that bridgestrap measures whether your obfs4 port is reachable, and it uploads these results to the metrics sites, and the metrics sites use them if available instead of looking at Running. But currently the "uploads to metrics" step happens once a day, while bridgestrap produces results way more frequently than that. And in the past there were surprises where the metrics side would say something like "I'll call you running if bridgestrap said you were reachable within the past three hours" -- which is a great design if you are getting the bridgestrap results rapidly, but not great if you get them once a day. So, I'm glad to see that your relay was green again for a bit, but I fear a bit more work remains until we get there *consistently*. :) --Roger

On 6/23/24 21:59, Roger Dingledine wrote:
On Sun, Jun 23, 2024 at 09:28:23PM +0200, lists@for-privacy.net wrote:
A few months ago there was a recommendation to not exposing OrPort for bridges. This had the unpleasant effect that all bridges were 'red' on Tor Metrics, even though they were running perfectly fine. I noticed yesterday after the meeting that everything is 'green' again. https://metrics.torproject.org/rs.html#search/ForPrivacyNET
Thank you, I believe these 6 people did that: https://gitlab.torproject.org/tpo/network-health/team/-/issues/318 Right -- I think we are still in the process of fixing that issue. The current situation as I understand it is that bridgestrap measures whether your obfs4 port is reachable, and it uploads these results to the metrics sites, and the metrics sites use them if available instead of looking at Running.
But currently the "uploads to metrics" step happens once a day, while bridgestrap produces results way more frequently than that.
And in the past there were surprises where the metrics side would say something like "I'll call you running if bridgestrap said you were reachable within the past three hours" -- which is a great design if you are getting the bridgestrap results rapidly, but not great if you get them once a day.
So, I'm glad to see that your relay was green again for a bit, but I fear a bit more work remains until we get there *consistently*. :)
Yes I think the issue is somewhat hidden into collector and how/how often the tests from bridgestrap are processed. I am counting to work on this when we are back the 2nd week of july. Cheers, -hiro
_______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays

On 6/26/24 16:46, Hiro wrote:
On 6/23/24 21:59, Roger Dingledine wrote:
On Sun, Jun 23, 2024 at 09:28:23PM +0200, lists@for-privacy.net wrote:
A few months ago there was a recommendation to not exposing OrPort for bridges. This had the unpleasant effect that all bridges were 'red' on Tor Metrics, even though they were running perfectly fine. I noticed yesterday after the meeting that everything is 'green' again. https://metrics.torproject.org/rs.html#search/ForPrivacyNET
Thank you, I believe these 6 people did that: https://gitlab.torproject.org/tpo/network-health/team/-/issues/318 Right -- I think we are still in the process of fixing that issue. The current situation as I understand it is that bridgestrap measures whether your obfs4 port is reachable, and it uploads these results to the metrics sites, and the metrics sites use them if available instead of looking at Running.
But currently the "uploads to metrics" step happens once a day, while bridgestrap produces results way more frequently than that.
And in the past there were surprises where the metrics side would say something like "I'll call you running if bridgestrap said you were reachable within the past three hours" -- which is a great design if you are getting the bridgestrap results rapidly, but not great if you get them once a day.
So, I'm glad to see that your relay was green again for a bit, but I fear a bit more work remains until we get there *consistently*. :)
Yes I think the issue is somewhat hidden into collector and how/how often the tests from bridgestrap are processed.
I am counting to work on this when we are back the 2nd week of july.
I do not want to declare victory too soon, but I think this issue should be resolved. There was a configuration option hidden in collector that was making it process bridgestrap tests every 8 hours. I have now changed it to every hour. Let's see. Cheers, -hiro
Cheers,
-hiro
_______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
_______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays

On Montag, 8. Juli 2024 16:49:04 CEST Hiro wrote:
I do not want to declare victory too soon, but I think this issue should be resolved. There was a configuration option hidden in collector that was making it process bridgestrap tests every 8 hours. I have now changed it to every hour.
This has been looking damn good for 4 days :-) https://metrics.torproject.org/rs.html#search/ForPrivacyNETbr Flags, Uptime and green dot is OK By the way: I saw that everything is being prepared to shut down bridgedb and migrate everything to rdsys. -- ╰_╯ Ciao Marco! Debian GNU/Linux It's free software and it gives you freedom!

On 7/12/24 00:52, boldsuck wrote:
This has been looking damn good for 4 days :-) https://metrics.torproject.org/rs.html#search/ForPrivacyNETbr Flags, Uptime and green dot is OK
But the behaviour is not fixed. It just happened less often. -- Toralf

On 7/29/24 19:50, Toralf Förster via tor-relays wrote:
On 7/12/24 00:52, boldsuck wrote:
This has been looking damn good for 4 days :-) https://metrics.torproject.org/rs.html#search/ForPrivacyNETbr Flags, Uptime and green dot is OK
But the behaviour is not fixed. It just happened less often.
Collector should process the information it receives from rdsys every hour, but there might be a gap between when onionoo process the data and receives the information from collector. I'll see if there is an easy fix to this and let everybody know. -hiro
-- Toralf
_______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
participants (5)
-
boldsuck
-
Hiro
-
lists@for-privacy.net
-
Roger Dingledine
-
Toralf Förster