On Aug 19, 2011, at 11:52 AM, Karsten Loesing wrote:

Hi Sambuddho,

On 8/13/11 12:11 AM, Sambuddho Chakravarty wrote:

...

I see some relays with uptimes over 100 days (eg. Pandora14), but they don't have the stable flag enabled ... In the FAQ page it says that relays are marked stable when they are running for more than a day, however in the directory service page I see relays which are running for more than 200 days but not marked stable and some which are running for about 3 hours already have the stable flag. Why is there this difference ?

Well, the general answer is that the Stable flag is based on a relay's uptime history, not on the relay's current uptime. dir-spec.txt says:

"Stable" -- A router is 'Stable' if it is active, and either its Weighted MTBF is at least the median for known active routers or its Weighted MTBF corresponds to at least 7 days. Routers are never called Stable if they are running a version of Tor known to drop circuits stupidly. (0.1.1.10-alpha through 0.1.1.16-rc are stupid this way.)

So, if a relay had a really good uptime history, went offline, and came back online 3 hours ago, it can still have the Stable flag.

But I admit that it seems unlikely that a relay with 100 or 200 days current uptime has an uptime history that doesn't justify the Stable flag.

If you look up the various votes for these relays, you'll see that the directory authorities have different opinions here:

https://metrics.torproject.org/consensus-health.html

I asked Jake who runs Pandora14 why it doesn't have the Stable flag. But we didn't come up with an easy answer. It could be a bug.

Do you want to help find the bug? What other relays did you find that should have the Stable flag?

Once we have enough information about the bug we should open a ticket:

https://trac.torproject.org/projects/tor

Thanks, Karsten

I don't think this is a bug. If a flaky relay is unavailable regularly to at least some dirauths, they will (rightfully so!) count the relay as down for that time. The relay itself doesn't note that it was unavailable, and doesn't reset its uptime to 0. It is quite possible that a relay that is - for example - close to the maximum number of connections all the time and thus fails new connections frequently, while already established connections keep working. The dirauths might count that as downtime. Especially for a relay that changes between the "works" and "doesn't work" states quickly a stable flag shouldn't be given out. In general, don't look at the uptime field in router descriptors. This value is set and tracked by the relays themselves and can be influenced arbitrarily.