You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current ReqResp implementation cycles through nodes serially, meaning that the timeouts are all dependent on each other. This is by design, as tx requests which often do not care which peer they are engaging with will require some hard timeout so that we do not stall the node.
ReqResp work should be non blocking and entirely asynchronous
Track how often nodes are asking for data
How frequently they are asking
How long do the responses take
If we are unable to serve the normal usecase with serial requests, then we will have to parallelize requests among multiple peers.
The text was updated successfully, but these errors were encountered:
## Overview
Rather than serially sendinging request response through a list of
peers, we instead sample from the peer set.
Attempting to connect to a node that we do not already have a connection
open with.
If we are unable to get a free node after a number of attempts, we
simply bite the bullet and dial a peer we are
already connected to.
The aim is for this to decrease experienced rate limiting from other
peers
part of #8458
Overview
As mentioned by @just-mitch in #8434
The current ReqResp implementation cycles through nodes serially, meaning that the timeouts are all dependent on each other. This is by design, as tx requests which often do not care which peer they are engaging with will require some hard timeout so that we do not stall the node.
If we are unable to serve the normal usecase with serial requests, then we will have to parallelize requests among multiple peers.
The text was updated successfully, but these errors were encountered: