I’ve seen a few posts here about the mysterious ‘my v3 onion is suddenly unreachable’ (including the Whonix.org v3 onion, which predated my time here), and thought I’d let you know of a known issue about v3 descriptors being re-published to the HSdirs.
I first saw this in OnionShare whilst working on persistent v3 onions there, and this led to it being reported upstream. But even have seen it after migrating a simple website to another server (and migrating its v3 keys) and being surprised it wouldn’t be reachable (but then did, by itself, hours later)
In all such cases, as seem to be the cases I’ve seen in this forum, the v3 onion magically starts working again all by itself later, without any other action taken by the admin. This is because some sort of TTL/counter expires and eventually leads to the v3 onion successfully republishing its descriptor to the HSdir.
A fix has been implemented (we know it fixes the OnionShare use case of re-publishing a v3 descriptor) and is expected to arrive in Tor 0.3.5 in December.
Links of interest:
https://github.com/micahflee/onionshare/issues/461
https://github.com/micahflee/onionshare/issues/677
(Note the Tor ticket: the devs there are preoccupied with the fact that it affects ‘scalability’ of clustered v3 onions, and generally have overlooked the fact that the bug actually occurs for basic single-server v3 onions (and only when ‘restarting’ or republishing the key). However, at least one Tor dev in that ticket is aware that it fixes republication of a pre-generated v3 onion key too (the OnionShare use case, and the scenario in which I think many ‘my v3 onion has disappeared!’ cases can be attributed… just my theory though.))