[subsystem-bench] Emulate latency for messages and requests from node #4627
Conversation
AndreiEres
added
R0-silent
|
The CI pipeline was cancelled due to failure one of the required jobs. |
| @@ -561,6 +583,7 @@ async fn emulated_peer_loop( | |||
| } | |||
| } | |||
| if let Some(message) = message { | |||
| // TODO: message.peer() is always None because it's a message from node | |||
There was a problem hiding this comment.
Yes but not in this PR
| if self.has_latency() { | ||
| let latency = self.emulate_latency(); |
There was a problem hiding this comment.
nit: we could have a get_latency() which returns a Option<tokio::time::Sleep> instead of these two functions and use if let Some(latency) = peer.get_latency() {...
| @@ -548,6 +566,10 @@ async fn emulated_peer_loop( | |||
| emulated_peer.rx_limiter().reap(size).await; | |||
| stats.inc_received(size); | |||
|
|
|||
| if peer_message.is_from_node() && emulated_peer.has_latency() { | |||
| emulated_peer.emulate_latency().await; | |||
There was a problem hiding this comment.
This will effectively stall the peer loop for up to whatever latency/2 is configured, meaning the peer also won't send messages to the node, even if we request it to from main thread. This latency needs to be emulated in the task that is supposed to send messages from the node to the peer, see proxy_send_request method.
There was a problem hiding this comment.
Changed it. Is that what you were thinking about?
|
@sandreim Could you look at the changed implementation? |
| if latency_ms > 0 && receiver.is_connected() { | ||
| emulate_latency(latency_ms).await; | ||
| } | ||
| tx_network.send_message_to_peer(&peer, message); |
There was a problem hiding this comment.
Why do we still want to send to a disconnected peer ? Wouldn't the assert in send_message_to_peer trigger if peer is not connected ?
AndreiEres
force-pushed
the
AndreiEres/subsystem-bench-latency
branch
from
7bbc00b to
2f8647f
Compare
Fixes #4611