When retrieval data transfer fails, deal remains in Ongoing state

[dirkmc]

This bug was observed during manual testing.

When trying to make a retrieval deal for a CID with a miner, there was an error saying that there is an existing retrieval deal with the same miner for the same piece CID that is still active. The error message says the deal state is DealStatusOngoing:

lotus client  retrieve --miner f1234 bafykbzacecd3a2hrmxiw4g22ixtwmkfwq32o4cpyzvrogr3rxpna2yazmjbo1 output
> Recv: 0 B, Paid 0 FIL, ClientEventOpen (DealStatusNew)
ERROR: retrieval failed: Retrieve failed: there is an active retrieval deal with peer 12D3KooWQrZnhRaBecSZYKR815zH65xkg1TvfyPWUGYwehHvVFK3 for payload CID bafykbzacecd3a2hrmxiw4g22ixtwmkfwq32o4cpyzvrogr3rxpna2yazmjbo1 (retrieval deal ID 1, state DealStatusOngoing) - existing deal must be cancelled before starting a new retrieval deal

However list-transfers shows that the deal is in the Completed state:

$ lotus client list-transfers --completed
Sending Channels
ID                   Status     Sending To   Root Cid     Initiated?  Transferred  Voucher
1616506229013813701  Completed  ...ehHvVEKB  ...yazmjbo1  Y           583.7MiB     ...qrw5nr23pp5wwjgg2ybwqmot4rjuvgs6xm"}}

Note that in lotus there is currently no way to list retrieval deals or cancel a retrieval deal: filecoin-project/lotus#5870

[kernelogic]

I also noticed, if I ctrl+C at client side while the transfer is executing, the miner will continue serve content to I don't know where. Have to manually cancel-transfer on the client side in order to stop the miner.