GitLab

As @sebastian reported by IRC, sbws logs Dec 22 21:41:13 v3bwfile[283101]: <WARNING> (MainThread) It is more than 50%.

That line doesn't make sense without the previous info one The difference between the sum of the last consensus weight [...].

We should either explain the warning line what's the 50% about or lower the log level of that line to info.

Lowering the log level to info will generate less email notifications when configuring them as in #40110 (closed).

A different issue is whether it is acceptable that gabelmoo's consensus weight sum is more than the 50% of the consensus. We added the warning in #28216 (closed) while working on #27339 (closed).

It looks like gabelmoo's had always greatest consensus weight sum (see for instance the total consensus weight graph back in 2018 when the log was created: https://metrics.torproject.org/totalcw.png?start=2018-10-01&end=2018-10-31), what it probably due gabelmoo's location (Germany). What i'm not sure is whether it was as much as the 50% before because relays were getting stuck in bandwidth partitions with torflow or we just didn't notice cause torflow wasn't logging that.

@sebastian reported that this warning is gone the 02 Jan 2022, so i'd suggest that we keep the warning for now, explaining what the 50% is about and open a different issue to track how often the 50% difference is reached and whether that's a sbws bug.

What do you think, @gk?