One unlucky path, and everything looked broken
Mars market mirrors
marshjhtog245vzjzcicnmv2ci6yljibvdm4pngq5kmkfvcutppboxad.onion
marsiujka6lrsaqpnxiwvknthhzsrlmq77mnl2fi62guc4lwxif65syd.onion
marsmtbwtxkohhpu34m4jkwcntian7n257wsex5tbkmsjdjrsz6me3yd.onion
Printed as supplied, in no order. Nothing here is watched or timed, so an address that opens is not proof of anything. More about the set
The address will not open. Ten minutes later, on the same machine, with nothing changed except time, it opens instantly. Nothing about the service changed in between.
What changed was the route. A connection to an onion service runs through a chain of relays chosen partly at random, and some chains are simply bad.
- Responsible
- The network. A route was assembled from volunteer relays and this particular assembly performed badly.
- Usually blamed on
- The market, for being down, or the address, for being wrong.
- What follows
- One action fixes most of it and takes a second. Everything else people try here is unnecessary.
Why paths vary this much
Relays are run by volunteers on hardware that ranges from serious to whatever was in a cupboard. Capacity, latency and reliability differ enormously across them. Your client picks a path weighted towards capacity, but the weighting is a tendency and not a guarantee.
A connection to an onion service is also longer than an ordinary Tor connection. Both sides build their own legs and they meet in the middle. That means more relays involved, which means more chances for one of them to be the weak link on the day.
The one thing to do
Ask for a new circuit for that site. Tor Browser has this in its menu and it takes a second. What it does is discard the current path and build a fresh one, which resolves this case immediately if this case is what you had.
Do it once. Then wait a moment and do it once more if needed. Doing it repeatedly and quickly is counterproductive, partly because building paths takes time and partly because it looks like exactly the behaviour that a rate limiter exists to stop.
How this distinguishes itself
| Result | What it narrows to |
|---|---|
| New circuit fixes it | It was this case, and the question is closed |
| New circuit does not, other onion sites are fine | Points at the service side rather than at your routing |
| New circuit does not, no onion site works | Points at your own connection or at the whole network |
| Ordinary web is also broken | Not a Tor question at all |
That table is the most useful thing on this page. Four outcomes, one action, and the ambiguity collapses in under a minute.
Why the market gets blamed instead
Because the market is the thing you were trying to reach, so it is the thing in your mind when the attempt fails. The route is invisible. Nothing on screen mentions it, no error message names a relay, and the failure is presented as a property of the destination.
This is the single most common wrong attribution in the whole subject. A large share of the people who believe they have watched a market go down have watched a circuit go bad.
What follows from getting it right
You stop concluding things about the market from single failed page loads. That matters, because those conclusions travel. Somebody says a market is down, other people repeat it, and within an hour there is a consensus built on a handful of unlucky routes.
It also stops the address hunt, which is the genuinely risky behaviour that follows from believing an address is dead.