Some relays get stuck in a low bandwidth authority bucket, and never get out. And their self-measurement isn't enough to unstick them (#22453 (moved)).
So we want a tool that sends traffic through relays with low consensus weight, ideally enough to guarantee them the Fast flag (at which point, the client bandwidth allocation and network measurement take over).
It would be nice to just send drop cells.
But I think we can do this using stem and tor and some URL downloading library.
Sticking this in "Torflow" because that's where it might end up.
But maybe it would be better in its own component eventually.
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information
Child items 0
Show closed items
No child items are currently assigned. Use child items to break down this issue into smaller parts.
Linked items 0
Link issues together to show that they're related.
Learn more.
But they didn't seem to make the relays repost their descriptors, for one of two reasons:
most slow relays really are that slow, and
This seems to be true for most relays below 1000 consensus weight.
we don't know if DirPort bytes count in rephist (they seem to in the code, but we're not sure).
I think they do count.
But they were delayed because relays take up to 4 hours to notice when these numbers have gone up, because that's the current NUM_SECS_BW_SUM_INTERVAL.
We're trying it on 100 of the slowest relays. I'll attach the selection script and trial list / script.
Unless we want to continue abusing the directory protocol, we will need to run a tor client to do this, or wait until stem gets ORPort support (#18856 (closed)).