Greetings!
This is to announce that the Tor Project network team will soon change how fallback directories are selected as we are about to update that list.
As a reminder, here is the fallback directory definition from the tor man page:
When tor is unable to connect to any directory cache for directory info (usually because it doesn’t know about any yet) it tries a hard-coded directory. Relays try one directory authority at a time. Clients try multiple directory authorities and FallbackDirs, to avoid hangs on startup if a hard-coded directory is down. Clients wait for a few seconds between each attempt, and retry FallbackDirs more often than directory authorities, to reduce the load on the directory authorities
Previously, we would ask this list for volunteers and then start a lengthy process over weeks to add/remove volunteers from an official list and then make sure the volunteered relays were matching a certain set of criteria.
This process was very time consuming for our small team and so we are making a change to be more efficient and avoid to slip on the updates like we did in the previous versions.
We will now select at random relays to become fallback directories that match those requirements:
- Fast flag
"Fast" -- A router is 'Fast' if it is active, and its bandwidth is either in the top 7/8ths for known active routers or at least 100KB/s.
- Stable flag
"Stable" -- A router is 'Stable' if it is active, and either its Weighted MTBF is at least the median for known active routers or its Weighted MTBF corresponds to at least 7 days.
- DirCache is set to 1 (default)
- As been around for at least 90 days.
The above corresponds to more than 4000 relays at the moment in the consensus and so we'll randomly pick 200 from those at each release so we can have regular rotation and shift load over time over the entire network.
Over time, we will remove or add more relays at each minor release if the set of fallback directories not working reaches a 25% threshold or more.
Finally, it is very possible that we'll change those requirements over time as we assess this change. We'll do our best to inform this list in time.
Thanks! Tor Network Team
On 4/7/21 9:04 PM, David Goulet wrote:
Over time, we will remove or add more relays at each minor release if the set of fallback directories not working reaches a 25% threshold or more.
In the past a fallback dir volunteer committed himself to have address and port stable for 2 years.
If a relay is now removed from the fallback directory list - how long shall the Tor relay operator wait before he can change the address and/or port? Technically speaking, how long does it usually takes till a Tor (browser) version containing the new fallback dir is spreaded out mostly?
-- Toralf
On 07 Apr (21:43:50), Toralf Förster wrote:
On 4/7/21 9:04 PM, David Goulet wrote:
Over time, we will remove or add more relays at each minor release if the set of fallback directories not working reaches a 25% threshold or more.
In the past a fallback dir volunteer committed himself to have address and port stable for 2 years.
If a relay is now removed from the fallback directory list - how long shall the Tor relay operator wait before he can change the address and/or port?
No more requirements of such anymore. By selecting Stable and relays that have been around for a while, the theory is that the majority of relays of the selected list will be stable as in same address and port.
I do hope that overall in the network most relays do not change port/address often as relay stability is pivotal to an healthy network. But if so, our monitoring of the fallback directories will trigger an alert once too many relays are not working anymore as fallback and so we'll issue a new list.
Technically speaking, how long does it usually takes till a Tor (browser) version containing the new fallback dir is spreaded out mostly?
It is quite fast on TB side, see the updated graph:
https://metrics.torproject.org/webstats-tb.html
Cheers! David
Hi,
On 8. Apr 2021, at 14:50, David Goulet dgoulet@torproject.org wrote: On 07 Apr (21:43:50), Toralf Förster wrote:
On 4/7/21 9:04 PM, David Goulet wrote:
Over time, we will remove or add more relays at each minor release if the set of fallback directories not working reaches a 25% threshold or more.
In the past a fallback dir volunteer committed himself to have address and port stable for 2 years.
If a relay is now removed from the fallback directory list - how long shall the Tor relay operator wait before he can change the address and/or port?
No more requirements of such anymore. By selecting Stable and relays that have been around for a while, the theory is that the majority of relays of the selected list will be stable as in same address and port.
I do hope that overall in the network most relays do not change port/address often as relay stability is pivotal to an healthy network. But if so, our monitoring of the fallback directories will trigger an alert once too many relays are not working anymore as fallback and so we'll issue a new list.
Do we have some metrics in place to spot whether this improves/degrades bootstrapping performance and how it develops over time?
Cheers Sebastian
tor-relays@lists.torproject.org