StateDownloader should retry immediately after receiving empty reply from peer #1077
Assignees
Milestone
Comments
|
Seems like we need to keep some sort of tracking around which peers do not have certain trie nodes so that we can get to a point where we know "no peers have this trie node". I believe this is needed for us to get out of a stalled state sync as noted in #1073 |
gsalgado
added a commit
to gsalgado/py-evm
that referenced
this issue
Jul 26, 2018
Keep track of which nodes we requested to every peer so that we know which ones they're missing when they reply with just a subset of what was requested. That way we can avoid re-requesting those missing nodes to peers that don't have them. #1073 Also, when a reply from a peer contains only a subset of what was requested, we immediately re-request the missing ones. Closes: ethereum#1077
gsalgado
added a commit
to gsalgado/py-evm
that referenced
this issue
Jul 26, 2018
Keep track of which nodes we requested to every peer so that we know which ones they're missing when they reply with just a subset of what was requested. That way we can avoid re-requesting those missing nodes to peers that don't have them. #1073 Also, when a reply from a peer contains only a subset of what was requested, we immediately re-request the missing ones. Closes: ethereum#1077
gsalgado
added a commit
to gsalgado/py-evm
that referenced
this issue
Jul 30, 2018
Keep track of which nodes we requested to every peer so that we know which ones they're missing when they reply with just a subset of what was requested. That way we can avoid re-requesting those missing nodes to peers that don't have them. #1073 Also, when a reply from a peer contains only a subset of what was requested, we immediately re-request the missing ones. Closes: ethereum#1077
gsalgado
added a commit
to gsalgado/py-evm
that referenced
this issue
Jul 30, 2018
Keep track of which nodes we requested to every peer so that we know which ones they're missing when they reply with just a subset of what was requested. That way we can avoid re-requesting those missing nodes to peers that don't have them. #1073 Also, when a reply from a peer contains only a subset of what was requested, we immediately re-request the missing ones. Closes: ethereum#1077
gsalgado
added a commit
that referenced
this issue
Jul 30, 2018
Keep track of which nodes we requested to every peer so that we know which ones they're missing when they reply with just a subset of what was requested. That way we can avoid re-requesting those missing nodes to peers that don't have them. #1073 Also, when a reply from a peer contains only a subset of what was requested, we immediately re-request the missing ones. Closes: #1077
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
What is wrong?
During a state sync, sometimes a peer won't have the trie nodes we requested, and so will send us an empty
NodeDatamsg. When that happens we just move on and those nodes are re-requested only after the timeout (20s) has passed.How can it be fixed
Whenever we get an empty
NodeDatafrom a peer, we should look up the nodes that were originally requested to that peer and re-request them to a different peer.The text was updated successfully, but these errors were encountered: