#8419: Bw Weight Failure Warns
-----------------------+----------------------------------------------------
Reporter: mikeperry | Owner: mikeperry
Type: defect | Status: new
Priority: normal | Milestone:
Component: Tor | Version:
Keywords: | Parent:
Points: | Actualpoints:
-----------------------+----------------------------------------------------
Description changed by mikeperry:
Old description:
> We're getting a few of these:
> Mar 06 22:55:01.000 [notice] Computed bandwidth weights for Case 3be (E
> scarce, Wee=1, Wmd == Wgd) with v10: G=1972445 M=834651 E=
> 962503 D=1236746 T=5006345
> Mar 06 22:55:02.000 [warn] Bw Weight Failure for Case 3b: Etotal
> 1679546.562000 != Mtotal 1660246.975500. G=1968169 M=826231 E=970
> 923 D=1241022 T=5006345. Wgg=0.711500 Wgd=0.214500 Wmg=0.288500
> Wme=0.000000 Wmd=0.214500 Wee=1.000000 Wed=0.571000
>
> Notice that the G, M, E, D values all change between when we compute the
> weights (the notice) and when we spot-check the resulting consensus (the
> warn). The total bandwidth (the T value) interestingly did *not* change..
> So it's almost like bandwidth just moved from one flag to another
> somehow.
>
> Are we changing our opinion on node flags somewhere in between? Is that a
> result of the Fast flag changes, or the unmeasured capping? Are those
> both already in effect?
>
> For an extra datapoint, this seemed to only start happening when the bw
> auths failed.
New description:
We're getting a few of these:
Mar 06 22:55:01.000 [notice] Computed bandwidth weights for Case 3be (E
scarce, Wee=1, Wmd == Wgd) with v10: G=1972445 M=834651 E=962503 D=1236746
T=5006345
Mar 06 22:55:02.000 [warn] Bw Weight Failure for Case 3b: Etotal
1679546.562000 != Mtotal 1660246.975500. G=1968169 M=826231 E=970923
D=1241022 T=5006345. Wgg=0.711500 Wgd=0.214500 Wmg=0.288500 Wme=0.000000
Wmd=0.214500 Wee=1.000000 Wed=0.571000
Notice that the G, M, E, D flag total values all change between when we
compute the weights (the notice) and when we spot-check the resulting
consensus (the warn). The total bandwidth (the T value) interestingly did
*not* change.. So it's almost like bandwidth just moved from one flag to
another somehow, during the consensus process.
Are we changing our opinion on node flags somewhere in between these two
calculations? Is that a result of the Fast flag changes, or the unmeasured
capping? Are those both already in effect?
For an extra datapoint, this seemed to only start happening when the bw
auths failed.
--
--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/8419#comment:1>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online