Opened 2 years ago

Last modified 18 months ago

#21697 new defect

torflow bwfiles maybe should be updated

Reported by: tom Owned by:
Priority: Medium Milestone:
Component: Core Tor/Torflow Version:
Severity: Normal Keywords:
Cc: mikeperry, teor Actual Points:
Parent ID: Points:
Reviewer: Sponsor:

Description

A while ago we updated the bwfiles with new data: https://gitweb.torproject.org/torflow.git/commit/?id=2c5faf72f7443cd68a68e490645c2449b611283c

Maybe this should be done again. However, I have a duplicate line in mine that probably indicates a bug that would need to be fixed first.

Current:

1 32M
7 16M
20 8M
37 4M
56 2M
72 1M
86 512k
93 256k
98 128k
99 64k
100 32k
.

New:

0 64M
1 64M
3 32M
11 16M
28 8M
53 4M
71 2M
85 1M
94 512k
97 256k
98 128k
99 64k
100 32k
.

Child Tickets

Change History (12)

comment:1 Changed 2 years ago by teor

Status: newneeds_review

I think the bug is in this line:
https://gitweb.torproject.org/torflow.git/tree/NetworkScanners/BwAuthority/aggregate.py#n349

There's no way we should be adding the same file when f is 0, and when f is 1.

Please see my branch bug21697 at https://github.com/teor2345/torflow.git for a fix.
(Or, to view the code: https://github.com/teor2345/torflow/tree/bug21697 )

But what I think this means is that some relays are getting really fast now, and we need to add a 128M file. I opened #21698 for this.

comment:2 Changed 2 years ago by teor

Oh, I should add that I tested this on a bandwidth authority on the tor test network.
I can't confirm the bug, because our speeds only get us as high as 4M.
But I can confirm that the change does not *delete* any of the existing entries.

comment:3 Changed 19 months ago by teor

This is in testing on the live network on a duplicate bandwidth authority.

comment:4 Changed 19 months ago by teor

bwfiles is automatically updated by aggregate.py, so when we update it in the repository, it's only used for initial bandwidth scanning (1 hour? 3 days?).

Still, it would be nice to have a file that starts at 64M.

comment:5 Changed 19 months ago by teor

Priorities and Severities in torflow are meaningless, setting them all to Medium/Normal.

comment:6 Changed 19 months ago by teor

Owner: aagbsn deleted
Status: needs_reviewassigned

aagbsn was the default owner, unassigning

comment:7 Changed 19 months ago by tom

Here's every bwauth's bwfile:

maatuska

1 16M
15 8M
48 4M
70 2M
85 1M
94 512k
97 256k
98 128k
99 64k
100 32k
.

maatuska duplicate

1 16M
14 8M
47 4M
70 2M
85 1M
94 512k
97 256k
98 128k
99 64k
100 32k
.

moria1

0 64M
1 32M
2 16M
21 8M
52 4M
72 2M
85 1M
94 512k
97 256k
98 128k
99 64k
100 32k

gabelmoo

1 32M
4 16M
13 8M
27 4M
42 2M
58 1M
74 512k
89 256k
97 128k

bastet

1 8M
13 4M
44 2M
67 1M
79 512k
87 256k
92 128k
94 64k
95 32k
.

faravahar (Singapore)

10 4M
54 2M
79 1M
92 512k
96 256k
98 128k
99 64k
100 32k
.

comment:8 Changed 18 months ago by tom

Prelim results: https://i.imgur.com/BvOifZG.png

Seems to not make a big difference, but I'll let it running over the weekend

comment:9 Changed 18 months ago by tom

Here's a better image: https://i.imgur.com/J7GJtlW.png

comment:10 in reply to:  8 ; Changed 18 months ago by teor

Replying to tom:

Prelim results: https://i.imgur.com/BvOifZG.png

Seems to not make a big difference, but I'll let it running over the weekend

Are you testing my bug21697 branch?
I don't expect it to make a difference.
Even when the bug is triggered, it should have no effect or a minimal effect on the results.

Or are you modifying the default bwfiles content?
To test that, you'll need a clean checkout with no previous data, or the first aggregate run will overwrite bwfiles.

Also, I can't read the labels on your graph.

comment:11 in reply to:  10 Changed 18 months ago by tom

Replying to teor:

Replying to tom:

Prelim results: https://i.imgur.com/BvOifZG.png

Seems to not make a big difference, but I'll let it running over the weekend

Are you testing my bug21697 branch?
I don't expect it to make a difference.
Even when the bug is triggered, it should have no effect or a minimal effect on the results.

Correct, this is what it's running.

comment:12 Changed 18 months ago by teor

Status: assignednew

Mark all tickets that are assigned to nobody as "new".

Note: See TracTickets for help on using tickets.