Accounting and the 'Stable' flag
Hello list, I have enabled daily accounting on a non-exit Tor relay some days ago. On the first day, the transfer threshold was reached and the relay entered hibernation for a few hours, as expected. On the next day, I reduced RelayBandwidthRate to ensure that the accumulated transfer would fall just short of the AccountingMax value. During the last days, the relay has only been offline once for a couple of minutes (software update), besides the previously mentioned single hibernation period. As of today, the relay seems to have lost the 'Stable' flag, which I don't really understand. Am I missing something here? I could not find documentation mentioning a relationship between accounting and the 'Stable' flag. -Ralph
On 21 Aug 2016, at 13:11, Ralph Seichter <tor-relays-ml@horus-it.de> wrote:
Hello list,
I have enabled daily accounting on a non-exit Tor relay some days ago. On the first day, the transfer threshold was reached and the relay entered hibernation for a few hours, as expected. On the next day, I reduced RelayBandwidthRate to ensure that the accumulated transfer would fall just short of the AccountingMax value.
During the last days, the relay has only been offline once for a couple of minutes (software update), besides the previously mentioned single hibernation period. As of today, the relay seems to have lost the 'Stable' flag, which I don't really understand.
Am I missing something here? I could not find documentation mentioning a relationship between accounting and the 'Stable' flag.
-Ralph
Any event where your relay is unreachable by the directory authorities gets counted as something that could influence your Stable flag. Also, the flag is given out in relation to the stability of the other relays. Just keep your relay up and you should regain the flag. Cheers Sebastian
On 21.08.2016 13:29, Sebastian Hahn wrote:
Just keep your relay up and you should regain the flag.
Followup: Now that another day has passed, the relay's traffic has unfortunately dropped significantly, to roughly 10% of what the relay could process according to AccountingMax and the specified bandwidth limitations. The number of connections has dropped to approximately 20% of the former value. That's quite a shame. Can I do anything to actively promote the relay, or am I left with leaving it running and hoping that it is not "tainted" in some way? The 'Fast' flag is still present, but actual use is minimal. -Ralph
@Ralph https://blog.torproject.org/blog/lifecycle-of-a-new-relay It is normal to experience a dip in traffic. Just keep the relay running and let things play out. It can take weeks for utilization to peak.
On 22.08.16 19:22, Green Dream wrote:
Thank you. I had read about this, but the relay in question already had both the 'stable' and 'guard' flags (which probably meant it was in phase 3), and lost both flags after I enabled accounting. I don't mean to be impatient, though, and I will keep the relay running and see if the coming week brings changes. I just want to make sure that a) I did not screw up configuring things and b) the ISP is not throttling traffic without notification. The latter possibility only occurred to me some hours ago. -Ralph
I just want to make sure that... the ISP is not throttling traffic without notification.
Yeah, that's a valid concern. Aside from running through some of the M-LAB tests (https://www.measurementlab.net/tests/), it can be hard to predict or detect this. Another issue that can crop up is poor peering capacity between your ISP and the rest of the Internet. In particular, in order to get a good consensus weight, the Tor bandwidth authorities need a good path to your relay. This can be really hard to measure. I have run into at least one provider that had great bandwidth when measured with nearby hosts or localized CDNs, but somewhere in the path to the bandwidth authorities there was congestion or some other issue preventing good connectivity.
participants (3)
-
Green Dream -
Ralph Seichter -
Sebastian Hahn