Miner should not try to dial the client to restart a data transfer

[dineshshenoy]

From @dirkmc

It appears that on restart the miner is trying to dial the client to restart a data transfer.
The miner shouldn't dial the client - the client should restart when the connection goes down.

error restarting data transfer: unable to send restart request: exhausted %!d(float64=5) attempts but failed to open stream, err: failed to dial 12D3KooWHAN67HgLoqrygQnouFF5uVhwM4kP7zurpR65Powtfgsa: all dials failed

• [/ip4/127.0.0.1/tcp/33695] dial tcp4 127.0.0.1:33695: connect: connection refused
• [/ip6/::1/tcp/39135] dial tcp6 [::1]:39135: connect: connection refused
• [/ip4/10.240.0.46/tcp/33695] dial tcp4 0.0.0.0:9999->10.240.0.46:33695: i/o timeout
• [/ip4/13.68.239.146/tcp/33695] dial tcp4 0.0.0.0:9999->13.68.239.146:33695: i/o timeout

Recommendation from @magik6k and @hannahhoward

1. Miner shouldn't try to restart the request.
2. Client should auto-retry when it sees the network connection drop (every 5 minutes)

[hannahhoward]

Follow-up for @dirkmc

Specs for this ticket:

• remove auto restart for data transfers from provider in go-fil-markets - just leave deal as is, in transferring state
• in go-fil-markets, when a transfer becomes stalled (see event / message) on the client only start a go routine with a timer that attempts to restart the data transfer every 5 minutes -- I would rather NOT put this in go-data-transfer itself - not all users (including, most importantly, the provider) will want to attempt to restart

[dineshshenoy]

The miner retry is causing the deal to fail:

From @magik6k

bafyreig53wbhyafwc4nvdnupjadb3r24v6xlvsizpgbnnays7zqaec5kzq 0 StorageDealError f3rab4mhsr7f2gdg6ozgj44mtny6bhcy3rvlftc5xyki6t4hvhdkaji3ahfaid6fh2o3wff4yfrzag4lge2r2q 8GiB 0.004161616 FIL 520202 error restarting data transfer: unable to send restart request: exhausted %!d(float64=5) attempts but failed to open stream, err: routing: not found


[kernelogic]

My client lotus freeze and it keep printing error messages like this. By "freeze" I mean the daemon still prints something, but any cli command for example lotus sync wait will freeze and cannot terminate by CTRL+C.

2020-11-27T08:07:27.307Z WARN storagemarket_network network/libp2p_impl.go:127 exhausted 5 attempts but failed to open stream, err: dial backoff
2020-11-27T08:07:27.307Z WARN storagemarket_impl clientstates/client_states.go:176 error when querying provider deal state: %!w(*xerrors.wrapError=&{failed to open stream to miner 0xc29c4054c0 {[9796235 25912814 25925750]}})
2020-11-27T08:07:27.372Z INFO markets loggers/loggers.go:14 storage event {"name": "ClientEventWaitForDealState", "proposal CID": "bafyreibglvcufodk3t4djxa37tsqamqdar27wnpqzbeo3npt3tyjsdfibq", "state": "StorageDealCheckForAcceptance", "message": ""}
2020-11-27T08:07:28.308Z INFO markets loggers/loggers.go:14 storage event {"name": "ClientEventWaitForDealState", "proposal CID": "bafyreierge755zb3wpx57o4mnhzsozf3bu3rpnaaidlbjvi7psozizevza", "state": "StorageDealCheckForAcceptance", "message": ""}
2020-11-27T08:08:13.273Z WARN storagemarket_network network/libp2p_impl.go:127 exhausted 5 attempts but failed to open stream, err: dial backoff
2020-11-27T08:08:13.273Z WARN storagemarket_impl clientstates/client_states.go:176 error when querying provider deal state: %!w(*xerrors.wrapError=&{failed to open stream to miner 0xc28c345000 {[9796235 25912814 25925750]}})
2020-11-27T08:08:14.273Z INFO markets loggers/loggers.go:14 storage event {"name": "ClientEventWaitForDealState", "proposal CID": "bafyreibsbbtv6a56coy7bybd6z33i73ro5a3ycnzyf3rkxaypt2rnj6upa", "state": "StorageDealCheckForAcceptance", "message": ""}