Increase base consensus timeout and make it a config #4532
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sui has a timeout on sending transaction to the nw consensus and automatically adjusts the timeout based on recent latency from nw it has seen in case of success, or linearly increases in case of timeout. In the stress test network we see quite a few of nw timeouts from the sui transaction submission side, and we indeed [see](http://grafana.shared.internal.sui.io:3000/goto/wqnUWbGVz?orgId=1) that control delay is adjusted accordingly to the timeouts and nw is not stuck. However, it seems that current timeout adjustment logic drops timeout pretty quickly which causes timeouts again. While this is not end of the world, it is noise and probably suboptimal. It also can be noticed that current base timeout value is very close to the actual consensus delay, so the timeout has to be frequently adjusted. Therefore, a short-medium term fix (this PR) is just to bump the delay step for timeout calculations so that it is further away from median nw latency. In the future we can improve timeout calculation by using weighted avg over a longer window instead of weighting current and next timing only.
lxfind
reviewed
Sep 8, 2022
| @@ -152,6 +152,7 @@ impl NodeConfig { | |||
| pub struct ConsensusConfig { | |||
| pub consensus_address: Multiaddr, | |||
| pub consensus_db_path: PathBuf, | |||
| pub delay_step: Option<u64>, | |||
There was a problem hiding this comment.
Alternatively I think it may be cleaner if you don't use Option here but provide a serde default?
There was a problem hiding this comment.
Oh, too late :)
I don't really have a strong opinion on that, can use default next time
There was a problem hiding this comment.
The benefit of serde default is that we can then deploy this without a new config file
There was a problem hiding this comment.
I think it is same with Option, it deserializes to None by default
longbowlu
pushed a commit
that referenced
this pull request
Sep 11, 2022
Sui has a timeout on sending transaction to the nw consensus and automatically adjusts the timeout based on recent latency from nw it has seen in case of success, or linearly increases in case of timeout. In the stress test network we see quite a few of nw timeouts from the sui transaction submission side, and we indeed [see](http://grafana.shared.internal.sui.io:3000/goto/wqnUWbGVz?orgId=1) that control delay is adjusted accordingly to the timeouts and nw is not stuck. However, it seems that current timeout adjustment logic drops timeout pretty quickly which causes timeouts again. While this is not end of the world, it is noise and probably suboptimal. It also can be noticed that current base timeout value is very close to the actual consensus delay, so the timeout has to be frequently adjusted. Therefore, a short-medium term fix (this PR) is just to bump the delay step for timeout calculations so that it is further away from median nw latency. In the future we can improve timeout calculation by using weighted avg over a longer window instead of weighting current and next timing only.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Sui has a timeout on sending transaction to the nw consensus and automatically adjusts the timeout based on recent latency from nw it has seen in case of success, or linearly increases in case of timeout.
In the stress test network we see quite a few of nw timeouts from the sui transaction submission side, and we indeed see that control delay is adjusted accordingly to the timeouts and nw is not stuck.
However, it seems that current timeout adjustment logic drops timeout pretty quickly which causes timeouts again.
While this is not end of the world, it is noise and probably suboptimal.
It also can be noticed that current base timeout value is very close to the actual consensus delay, so the timeout has to be frequently adjusted. Therefore, a short-medium term fix (this PR) is just to bump the delay step for timeout calculations so that it is further away from median nw latency.
In the future we can improve timeout calculation by using weighted avg over a longer window instead of weighting current and next timing only.