bug: sync stuck due to networking error

[hunjixin]

reproduce on mocha network

1. setup a new full node to sync,
2. close the network(turn off wifi, pull out wire, etc)
3. wait for all peers to fail
4. open the network
5. the sync goroutine will never recover

the sync goroutine is stucking

if network is down, the peers in peerQuene will always be in a decreasing state due to errors that not a ErrNotFound one(network connect fail etc). and eventually run out of the havePeer channel. in this time GetRangeByHeight alway wait for hasPeer channel while getRangeByHeight wait for the result channel

a candidate fix is push back the peer state for errEmptyResponse error. #238

but another fix is to add timeout for GetRangeByHeight

func (s *Syncer[H]) requestHeaders(
	ctx context.Context,
	fromHead H,
	to uint64,
) error {
	amount := to - fromHead.Height()
	// start requesting headers until amount remaining will be 0
	for amount > 0 {
		size := header.MaxRangeRequestSize
		if amount < size {
			size = amount
		}

		to := fromHead.Height() + size + 1
		s.metrics.rangeRequestStart()
                 //to fix , add timeout for this context
		headers, err := s.getter.GetRangeByHeight(ctx, fromHead, to)
		s.metrics.updateGetRangeRequestInfo(s.ctx, int(size)/100, err != nil)
		s.metrics.rangeRequestStop()
		if err != nil {
			return err
		}

		if err := s.storeHeaders(ctx, headers...); err != nil {
			return err
		}

		amount -= size // size == len(headers)
		fromHead = headers[len(headers)-1]
	}
	return nil
}

[Wondertan]

Hey @hunjixin, thanks for detailed bug report. I am gonna take a look soon, unless @vgonkivs wants to beat me to it and look faster

[vgonkivs]

Thanks for opening an issue @hunjixin. I will take a closer look and report you back asap.