Skip to content
v2.1These docs are for deaddrop 2.1: serve local files, stdin and URLs, allow-listed clients, expiring tickets, resumable fetches and seven new notifiers.What's new in 2.1 →

Troubleshooting

Messages below are quoted as deaddrop prints them on stderr. For more detail, turn logging on with RUST_LOG=info (see Progress & logging).

A 2.1 server says why it won’t serve a client. Only a busy slot is retried; everything else stops at once.

“another client is downloading from this single-use server”

Section titled ““another client is downloading from this single-use server””

Someone else got the single-use slot first. fetch waits quietly for up to --retry-for seconds (default 120) in case that download fails and the slot frees up.

  • If it then gives up: run the same command again later, or ask the sender whether someone else already has it.
  • If you see “the server was busy with another client’s download and then stopped answering”, that other download most likely finished and the server deleted the data and shut down. Ask for a new ticket.
  • If you see “the server’s download slot is still taken, probably by this client’s dropped connection”, it’s your own earlier connection. After a crash or a lost network, the server takes up to about 30 s to notice. Just wait; fetch keeps retrying.

“this single-use drop has already been downloaded”

Section titled ““this single-use drop has already been downloaded””

Someone (maybe you, in an earlier run) already fetched it, and the server deleted the data. Ask the sender for a new ticket.

“this ticket expired before it was downloaded”

Section titled ““this ticket expired before it was downloaded””

The server was started with --expire-after and the time ran out. The data is gone. Ask for a new ticket.

“this ticket no longer works: the server was reset to serve something else”

Section titled ““this ticket no longer works: the server was reset to serve something else””

Someone ran deaddrop reset. Ask for the new ticket.

“the server isn’t serving this ticket”

Section titled ““the server isn’t serving this ticket””

Usually a ticket that was cut off when it was copied, or an old one. Tickets are long; check you have the whole line.

“the server only accepts specific clients”

Section titled ““the server only accepts specific clients””

The server uses --allow and your id isn’t on the list. The message tells you what to do:

  • If it shows an id, send that id to the server’s operator to add with --allow <id>.
  • If it says “this fetch had no identity”, run deaddrop id first, send the id it prints, and fetch again once they’ve added it.

If you were allowed before, check you’re using the same identity file (or that nobody ran deaddrop id --new).

The server runs a version older than 2.1, which doesn’t say why. It’s one of the reasons above (busy, already downloaded, or reset). Ask the sender to upgrade.

“the partial download is kept in .deaddrop-…; run the same command again to resume”

Section titled ““the partial download is kept in .deaddrop-…; run the same command again to resume””

The connection dropped and fetch gave up after --retry-for seconds without progress. Nothing is lost: run the same command with the same -o and it only downloads what’s missing. Delete the .deaddrop-<hash> folder to give up.

On a slow or flaky link, give it more time: --retry-for 600 (or retry_for under [fetch]).

“deleted the partial download; it can’t be finished”

Section titled ““deleted the partial download; it can’t be finished””

The server said the drop was already downloaded, expired or replaced, so what you had can never be completed. Ask for a new ticket.

The .deaddrop-<hash> folder is still there after an error

Section titled “The .deaddrop-<hash> folder is still there after an error”

That’s on purpose: it holds the partial download (or, for encrypted data, the whole download waiting for the key). Run the same command again, or delete it.

“the data is encrypted: pass the decryption key”

Section titled ““the data is encrypted: pass the decryption key””

The server used --encrypt. Run the same command again with DEADDROP_KEY=AGE-SECRET-KEY-1.... The download is kept, so this works without the server, even after a single-use server has shut down.

“note: the data is not encrypted; ignoring the decryption key”

Section titled ““note: the data is not encrypted; ignoring the decryption key””

Harmless: the server didn’t use --encrypt, so the key isn’t needed.

An error, and no file is written under its real name. Check you have the key that was printed with this ticket: every torrent, and every reset, gets a new one.

Since 2.1, a single-use server deletes its data once the download is done, also from a --data-dir, and when a ticket expires. Use --multi-use if the data should stay. Local files you serve directly (deaddrop serve ./file) are never deleted.

“–name only applies to stdin (-) or an http(s) download”

Section titled ““–name only applies to stdin (-) or an http(s) download””

--name sets the file name for - or a URL. Local files and torrents keep their own names.

Section titled ““–as-is can’t serve a magnet link; it has no file to serve””

--as-is serves a .torrent file or URL as a plain file. A magnet link has nothing to serve as a file.

A directory with no regular files in it. Symlinks inside directories, special files and empty folders are skipped.

Local files can’t be inside the data dir

Section titled “Local files can’t be inside the data dir”

deaddrop refuses to serve local files from inside its own data dir, since it deletes data there. Move them, or pick another --data-dir.

The client rejects a download of local files

Section titled “The client rejects a download of local files”

The files changed while they were served. The client checks every chunk against the hash taken when serving started. Don’t change files while they’re served; restart serve after changing them.

“bad –expire-after” / “bad –allow”

Section titled ““bad –expire-after” / “bad –allow””

--expire-after takes durations like 90s, 30m, 24h, 7d or 1h30m (a bare number is seconds). --allow takes the 64-character hex id that deaddrop id prints.

The server prints notify: warning: <notifier> failed: <error> and keeps serving; the ticket is still on stdout. Common causes:

  • Email: a provider that wants an app password, @ or : in the user name or password that isn’t percent-encoded, or the wrong scheme for the port (smtps:// for 465, smtp:// for 587).
  • Twilio: a trial account texting a number that isn’t verified, or a number that isn’t in E.164 form (+15552223333).
  • Telegram: the bot isn’t in the chat (or isn’t an admin of the channel), or the chat id is missing its -100 prefix.
  • ntfy: a protected topic without token or user:pass@.

Setting up notifiers walks through each one, with a way to send a test message.

“email needs a from address” / “email needs at least one to address”

Section titled ““email needs a from address” / “email needs at least one to address””

Add ?from=…&to=… to the URL, or from and to in its [[serve.notify]] table.

With --hide-key, the key is never printed, so at least one notifier must have include_key (or +key). Otherwise nobody would get it.

If fetch can’t connect at all, check the path it prints, and see Connections & relays. With relays disabled on either side, both must reach each other directly.