On 2012-09-11 22:43 , torhoste@tor.host-ed.me wrote:

Hi,

this is my very first post to this list. I operate tor exit relays vks24784 and vks24949 and subscribed to Tor Weather Reports. Now, at Sun, September 2, 2012 12:17 am and Thu, September 6, 2012 12:33 am, I received "Node Down" Reports for either of the nodes, 1st report for vks24949, 2nd for vks24784. That said, I noticed that these events reset the respective uptime counters at Atlas web application to "zero"

Uptime is for the Tor binary. Thus if the Tor binary restarts uptime will go back to 0.

(And thus when making config changes it is suggested to do a reload instead; of course for version updates you need to restart it and thus when tracking git you'll always keep a low uptime)

and the nodes would lose the already earned stable, guard, named, etc. flags for a certain time period.

They keep those flags as long as the node comes back quick enough. (I am not aware of what the value of 'quick enough' is, could be a day, could be less though).

[..]

I even checked the "uptime" through ssh and there hasn't been any reboots occurring.

But what about the tor binary itself?

Do a 'ps -ef | grep tor' and you will see the start time (STIME).

Also, I'm wondering what actually would happen to guard, stable, named, etc. flags if I'd actually need to reboot the nodes cause of maintenance?

Unless you keep the node offline for a longer time you will keep them.

I feel it unfortunate to lose these flags (also for the Tor network itself(?)) and would like to ask:

1. What does Tor Weather consider a "Node Down" event?

Please see:

https://gitweb.torproject.org/weather.git/blob_plain/HEAD:/doc/design.txt

Greets, Jeroen