Endorsements problems #1059

Closed massa-bot closed 2 years ago

In GitLab by @damip

End to end testing with 2 nodes shows multiple problems with endorsements:

blocks do not include all 8 endorsements as expected
cliques fork after a while

See attached logs of the 2 nodes.

logs1.xz

logs2.tar.xz

TODO: figure out where the problem(s) lie and correct them

In GitLab by @damip

changed the description

In GitLab by @sebf

changed the description

In GitLab by @damip

changed the description

In GitLab by @damip

assigned to @g-massa and @adolo

In GitLab by @damip

changed the description

In GitLab by @qdr

It seems to me that the endorsements of block qy9NVefPyWbk75FqS4cZ4PUx2eXPGQdvy4RKk2Sf9E4YJhuV7 are received before the block itself. See line 56139 of logs2.

In GitLab by @damip

@qdr that is not necessarily a problem I think, and may happen if address A is selected to create both a block and endorsements for that block: the block and the endorsements will be sent "simultaneously".

In GitLab by @AureliaDolo

I don't see the line where it should be tested that the endorsed block is the parent in the same thread

In GitLab by @qdr

@damip yeah you are right, there are examples where it happens and does not increase the number of cliques.

In GitLab by @damip

@adolo lines 904 - 917 in protocol_worker.rs

And to make sure that the stated parent is in the slot the endorsement pretends it is => lines 2582-2588 of block_graph.rs

In GitLab by @AureliaDolo

thanks :)

In GitLab by @AureliaDolo

After some end to end tests, everything seems to have been fixed by !154 :)

In GitLab by @AureliaDolo

solved

massalabs / massa

Endorsements problems #1059