A0-4346: Gather signatures and actually submit scores to chain #1907
Conversation
timorleph
force-pushed
the
A0-4346-bundesbrief
branch
from
478d4f5 to
1ab088b
Compare
timorleph
force-pushed
the
A0-4346-bundesbrief
branch
from
1ab088b to
e694d21
Compare
| debug!(target: LOG_TARGET, "Received ABFT score: {:?}.", score); | ||
| self.metrics.report_score(score[self.my_index]); | ||
| // TODO(A0-4339): sometimes submit these scores to the chain. | ||
| self.metrics.report_score(points[self.my_index]); |
There was a problem hiding this comment.
I guess self_my_index cannot ever be out of points bounds right?
There was a problem hiding this comment.
Yeah, it should always return points of the same length, equal to the total number of nodes. We could try handling failures there here, but I don't think it's worth it.
| session_over = true; | ||
| }, | ||
| }, | ||
| maybe_performance_hash = performance_from_scorer.next(), if !no_more_performance && !session_over => match maybe_performance_hash { |
There was a problem hiding this comment.
Do we need here && !session_over? What happens if we remove this?
There was a problem hiding this comment.
Not strictly. If it wasn't here, it would try submitting performance that is from the previous session. This will also work, by just failing at a later point, but imo it's clearer to already stop here – if you think otherwise I can remove this check.
| }; | ||
|
|
||
| const SCORE_SUBMISSION_PERIOD: usize = 300; |
There was a problem hiding this comment.
Will this aleph-node version publish some scores actually? I run it locally with below diff included, waited 300s, and nothing happened in the logs, nor in metrics
[14:55] marol-Latitude-5521:aleph-node ((e694d21b) *%) | for port in 9615 9616 9617 9618; curl localhost:9616/metrics 2>/dev/null | grep
"my_abft_score{"; end
my_abft_score{chain="a0dnet1"} 0
my_abft_score{chain="a0dnet1"} 0
my_abft_score{chain="a0dnet1"} 0
my_abft_score{chain="a0dnet1"} 0
There was a problem hiding this comment.
No because the first nonce in every session is 0 and should be 1.
There was a problem hiding this comment.
Also, remember that you have to specify finality version as current.
There was a problem hiding this comment.
I checked properly and it seems to be working as intended now. Still requires bumping finality version to 5, but this is by design.
There was a problem hiding this comment.
You mean bumping in primitives? There is a flag in run_nodes for that now.
Description
Send performance scores to the aggregator for gathering multisignatures under them and then submits them on chain. This should finish the ABFT performance project on the node side, modulo testing.
Type of change
Checklist: