Outcome misclassification on two storage deals

[raulk]

The following UUIDs are deals that concluded prematurely in the StorageDealProposalAccepted stage, and were marked as succeeding by the dealbot. Please analyse:

• 78d87834-5371-4dfd-91b1-d4dfcf8872f4
• c41c4751-a6f2-4d48-aa39-c03d4f95a41a

[hannahhoward]

My analysis is as follows:

• whether or not a deal is marked as "success" or "failed" is based on whether strorageDealExecutor.executeAndMonitorDeal returns an error
• there are two code paths by which we could return no error from this function
  • the first is when we hit StorageDealActive (https://github.com/filecoin-project/dealbot/blob/main/tasks/storage_deal.go#L390)
  • the second is when the deal update channel from ClientGetDealUpdates closes (https://github.com/filecoin-project/dealbot/blob/main/tasks/storage_deal.go#L346)

I am assuming something happened here that caused the second code path to get triggered. I think if the channel with Lotus closes for some reason we probably should mark the deal failed). However, I'd like to at least have a hypothesis as to why that channel could close -- network error? Lotus shutting down? I'm not sure. Either way I think we need to make line 346 return an error? I think?

[willscott]

we fixed the bug behind these. closing this issue