[metrics-team] bridge count: metrics.tpo vs. onionoo.tpo
karsten at torproject.org
Mon Sep 11 18:18:45 UTC 2017
On 2017-09-11 19:42, nusenu wrote:
> Hi Karsten,
> Karsten Loesing:
>> By default, Onionoo includes bridges that are not currently running but
>> that have been running in the past week. The idea is that relay/bridge
>> operators should be able to find their relay/bridge even when it's
>> offline for a short moment.
> yes I was also filtering for running bridges only (after parsing the
> entire details file), but it gets even more exciting here:
> "relays_published":"2017-09-11 16:00:00",
> "bridges_published":"2017-09-11 14:43:34",
> --> 3650 running.
> my python script says 5462 are running (mismatch 5462 vs 3650)
> so I cross checked this with a very simple grep:
> Download _all_ bridge entries (no running filter):
> grep ',"running":true,"' details.json |wc -l
> (note: the json file contains only briges 0 relays, otherwise this grep
> would give you incorrect results due to relay entries also matching)
> also: downloading all running bridges (no limit parameter) records gives
> you 3647 (not 3650) bridges.
> Is there anything wrong with that simple grep (and my python) or is
> onionoo's running=true wrong?
Huh, I think you found a bug.
It might be that we're using two different definitions for "running":
- The "running=false" parameter checks whether a bridge is contained in
the latest known status *and* has the "Running" flag assigned.
- The "running" field in details documents only checks whether a bridge
is contained in the latest known status but ignores whether it got the
"Running" flag there or not.
We should probably use the first definition and also require the
"Running" flag before calling a bridge "running".
Let's move this issue to Trac.
All the best,
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 495 bytes
Desc: OpenPGP digital signature
More information about the metrics-team