๐ฉบ doctor
One command that says what is wrong and which command fixes it.
pcloud doctorpCloud doctor
โ 1 problem found
Sync pair #1 ยท stuck
โ pcloud sync clear-tasks 1
Credential session token ยท 28 days left
API 21 of 22 endpoints reachable
Sync 5 pairs ยท daemon running
Sync โ 1 pair needs attention
โ #1 ~/pCloud/Documents โ /Documents
โ 1 queued operation with no destination
ยท report.tsv local id 9524
Not offered by pCloud
ยท list-rewind / restore-rewind
Expected โ `pcloud rewind` reconstructs it instead.What it checks
Your credential โ which tier you are on, and how long a session token has left. Session tokens expire after 30 days, or 7 unused.
Every endpoint the CLI calls โ by probing each without arguments. A method that exists gets far enough to complain about a missing parameter; one that does not 404s at the router first. So the error distinguishes "wrong call" from "no such thing", and nothing is performed.
The local sync daemon โ pair health read from pCloud Drive's own database, including the individual files stuck in a queue.
The verdict comes first
What is wrong opens the report, with the command that fixes it. Everything below
is evidence for that. Passing endpoints are a count rather than twenty lines of
green ticks โ --verbose lists them all.
Rewind is not a fault
list-rewind and restore-rewind call listrewindevents, which pCloud has
never exposed. It is reported under Not offered by pCloud rather than as a
failure, because pcloud rewind already reconstructs the same outcome from the
change log, the trash and per-file revisions.
Flagging a permanent, worked-around limitation as a fire on every single run teaches you to ignore the glyph that matters.
The remedy differs by fault
doctor names the specific command, because picking the wrong one is expensive:
| Fault | Remedy |
|---|---|
| Pair pointing at a deleted remote | pcloud sync prune <id> โ unpairs the folder |
| Healthy pair with a stuck queue | pcloud sync clear-tasks <id> โ clears the rows only |
| Endpoints needing a session token | pcloud login --session |
prune deletes around a thousand rows and unpairs the folder outright. That is
right for a zombie pair and catastrophic for a working one carrying two stale
queue entries, which is why doctor tells you which you have rather than leaving
you to guess.
No credential required
The local half runs without logging in โ a broken sync pair is exactly the kind of fault you hit before getting round to authenticating, so only the API half is given up on.