[tor-relays] Alleged Family Members / Stable flag "issues"

[Stijn Jonker]

Hi all,

For a little short of a year I'm running Relay SJC01 (328E54981C6DDD7D89B89E418724A4A7881E3192), there was some unnoticed outage of the relay which caused a couple days of downtime. This was at the end of Nov, oddly enough I don't seem to get back the "Stable" flag.

In searches I found some conflicting answers, it's either 7 days of uptime, a median of seven days of the entire uptime and/or the advice to check https://consensus-health.torproject.org/consensus-health-2017-12-14-18-00.html#328E54981C6DDD7D89B89E418724A4A7881E3192

What I don't understand is the differences in the output of the concensus:
- The concensus nodes that don't have IPv6 (assumed from "KnownFlags" from top of page. (longclaw, dizum, moria1 and faravahar) list my relay with Stable/Guard.
- The concensus nodes that do assign the "ReachableIPv6" flag, don't have my relay listed with those flags.

To test the IPv6 function, I took an VM outside of my network and ran Tor with "UseBridges" only allowing via iptables IPv6 out to my relay as entry node, and it the relay is/was functioning on IPv6.

Now happy to wait an other week/month etc for the stable flag. It doesn't add value to me, but I'm more curious why the flag doesn't return.

The same with the Alleged Family Member, I had an relay / exit with a reduced exit policy ($F8333E028E952840C1B93DAEE20880F75B90A68A) but that has been gone for quite some weeks (months) as it was taken down by the hoster. Now all info I can find is that when the relay doesn't announce this anymore one can expect it to be gone after some days/weeks. But in this case it's not going away.

If there are actions on my side happy to take them, but I can't find the "right thing" to do here.

Thanks!

[tor]

Post by Stijn Jonker
The same with the Alleged Family Member, I had an relay / exit with a reduced
exit policy ($F8333E028E952840C1B93DAEE20880F75B90A68A) but that has been gone
for quite some weeks (months) as it was taken down by the hoster. Now all
info I can find is that when the relay doesn't announce this anymore one can
expect it to be gone after some days/weeks.

F8333E028E952840C1B93DAEE20880F75B90A68A does seem to be gone from Atlas when searching for it directly. Have you removed the reference to it from the MyFamily line of the relay that's still up?

[Stijn Jonker]

Hi Tor,

Post by tor

Post by Stijn Jonker
The same with the Alleged Family Member, I had an relay / exit with a reduced
exit policy ($F8333E028E952840C1B93DAEE20880F75B90A68A) but that has been gone
for quite some weeks (months) as it was taken down by the hoster. Now all
info I can find is that when the relay doesn't announce this anymore one can
expect it to be gone after some days/weeks.

F8333E028E952840C1B93DAEE20880F75B90A68A does seem to be gone from
Atlas when searching for it directly. Have you removed the reference
to it from the MyFamily line of the relay that's still up?

Yes this entry is gone, shortly (believe 1 day) after the node was taken
down. Maybe it's due to the fact I comments the MyFamily config option
out, I'll put it back, just referencing itself. Maybe that will clear
things out.

Again not a major issue in my view, but it kind of feels like stale data
somewhere. If it's not on my side then I'm sure it will correct at some
point in time.

thx,
Stijn

[tor]

Yes this entry is gone, shortly (believe 1 day) after the node was taken down.
Maybe it's due to the fact I comments the MyFamily config option out, I'll put
it back, just referencing itself. Maybe that will clear things out.

Yep, I think that will probably fix it. You'll need to trigger Tor to re-upload its descriptor too. A restart of the Tor service will do it (although there may be a better way). I've noticed similar issues before with Atlas holding on to old nodes until that's done.

Cheers.

[teor]

Post by Stijn Jonker
For a little short of a year I'm running Relay SJC01 (328E54981C6DDD7D89B89E418724A4A7881E3192), there was some unnoticed outage of the relay which caused a couple days of downtime. This was at the end of Nov, oddly enough I don't seem to get back the "Stable" flag.
In searches I found some conflicting answers, it's either 7 days of uptime, a median of seven days of the entire uptime and/or the advice to check https://consensus-health.torproject.org/consensus-health-2017-12-14-18-00.html#328E54981C6DDD7D89B89E418724A4A7881E3192

The exact figure depends on each authority and its history of your
relay's stability. And how that stability compares to all other relays
it has measured.

Post by Stijn Jonker
- The concensus nodes that don't have IPv6 (assumed from "KnownFlags" from top of page. (longclaw, dizum, moria1 and faravahar) list my relay with Stable/Guard.
- The concensus nodes that do assign the "ReachableIPv6" flag, don't have my relay listed with those flags.

Does your relay have a stable IPv6 connection?
Was it down over IPv6 at some point in the past?

Alternately, the authorities that measure IPv6 may see the entire set of relays
as being less stable. (I can't imagine how they would see them as more stable.)
But if this were the case, they would be more likely to give the flag.

Post by Stijn Jonker
To test the IPv6 function, I took an VM outside of my network and ran Tor with "UseBridges" only allowing via iptables IPv6 out to my relay as entry node, and it the relay is/was functioning on IPv6.

How often does your IPv6 go down?

Post by Stijn Jonker
Now happy to wait an other week/month etc for the stable flag. It doesn't add value to me, but I'm more curious why the flag doesn't return.

Please wait a few more days.

T

[Stijn Jonker]

Hi Teor,

Post by teor

Post by Stijn Jonker
For a little short of a year I'm running Relay SJC01
(328E54981C6DDD7D89B89E418724A4A7881E3192), there was some unnoticed
outage of the relay which caused a couple days of downtime. This was
at the end of Nov, oddly enough I don't seem to get back the "Stable"
flag.
In searches I found some conflicting answers, it's either 7 days of
uptime, a median of seven days of the entire uptime and/or the advice
to check
https://consensus-health.torproject.org/consensus-health-2017-12-14-18-00.html#328E54981C6DDD7D89B89E418724A4A7881E3192

The exact figure depends on each authority and its history of your
relay's stability. And how that stability compares to all other relays
it has measured.

Post by Stijn Jonker
- The concensus nodes that don't have IPv6 (assumed from "KnownFlags"
from top of page. (longclaw, dizum, moria1 and faravahar) list my
relay with Stable/Guard.
- The concensus nodes that do assign the "ReachableIPv6" flag, don't
have my relay listed with those flags.

Does your relay have a stable IPv6 connection?
Was it down over IPv6 at some point in the past?

The IPv6 connectivity is not less stable then the IPv4. The IPv6 is
native, and used in day-2-day usage as well. Also smokeping/Nagios (ran
via an other device, but same uplink) don't report any differences in
connectivity (or the lack thereof).

Post by teor
Alternately, the authorities that measure IPv6 may see the entire set of relays
as being less stable. (I can't imagine how they would see them as more stable.)
But if this were the case, they would be more likely to give the flag.

Post by Stijn Jonker
To test the IPv6 function, I took an VM outside of my network and ran
Tor with "UseBridges" only allowing via iptables IPv6 out to my relay
as entry node, and it the relay is/was functioning on IPv6.

How often does your IPv6 go down?

On average, I would say one or two hours a month.

Post by teor

Post by Stijn Jonker
Now happy to wait an other week/month etc for the stable flag. It
doesn't add value to me, but I'm more curious why the flag doesn't
return.

Please wait a few more days.

:-) I'll wait thanks for the response.

Stijn