Client should stop polling provider for deal acceptance after (deal start epoch + grace period) has elapsed

[dirkmc]

Description

When making a storage deal, once the data has been successfully transferred to the provider, the client polls the provider to check on the state of the deal until the deal has been published:

go-fil-markets/storagemarket/impl/clientstates/client_states.go

Lines 171 to 209 in d5b1990

	// CheckForDealAcceptance is run until the deal is sealed and published by the provider, or errors
	func CheckForDealAcceptance(ctx fsm.Context, environment ClientDealEnvironment, deal storagemarket.ClientDeal) error {
	dealState, err := environment.GetProviderDealState(ctx.Context(), deal.ProposalCid)
	if err != nil {
	log.Warnf("error when querying provider deal state: %w", err) // TODO: at what point do we fail the deal?
	return waitAgain(ctx, environment, true, storagemarket.StorageDealUnknown)
	}
	if isFailed(dealState.State) {
	return ctx.Trigger(storagemarket.ClientEventDealRejected, dealState.State, dealState.Message)
	}
	if isAccepted(dealState.State) {
	if *dealState.ProposalCid != deal.ProposalCid {
	return ctx.Trigger(storagemarket.ClientEventResponseDealDidNotMatch, *dealState.ProposalCid, deal.ProposalCid)
	}
	return ctx.Trigger(storagemarket.ClientEventDealAccepted, dealState.PublishCid)
	}
	return waitAgain(ctx, environment, false, dealState.State)
	}
	func waitAgain(ctx fsm.Context, environment ClientDealEnvironment, pollError bool, providerState storagemarket.StorageDealStatus) error {
	t := time.NewTimer(environment.PollingInterval())
	go func() {
	select {
	case <-t.C:
	_ = ctx.Trigger(storagemarket.ClientEventWaitForDealState, pollError, providerState)
	case <-ctx.Context().Done():
	t.Stop()
	return
	}
	}()
	return nil
	}

If there is an error polling the provider, the client waits for 30 seconds then tries again.

go-fil-markets/storagemarket/impl/client.go

Line 45 in d5b1990

const DefaultPollingInterval = 30 * time.Second

If there is no error but the deal hasn't been accepted yet, the client also waits for 30 seconds then tries again.

Currently there is no upper bound on the number of retries.

Proposed solution

Keep polling until the deal proposal start epoch has been reached (plus 10 epochs).
At that point, fail the deal.

Increase polling interval to 5s (at the moment it's set to 1s in lotus)

[s0nik42]

Hi @dirkmc , Does the maximum number of error is also going to work for offline deals ?